dynamic analysis and fault detection of...
TRANSCRIPT
DYNAMIC ANALYSIS AND FAULT
DETECTION OF MULTI CRACKED
STRUCTURE UNDER MOVING MASS
USING INTELLIGENT METHODS
Shakti Prasanna Jena
Department of Mechanical Engineering
National Institute of Technology Rourkela
DYNAMIC ANALYSIS AND FAULT
DETECTION OF MULTI CRACKED
STRUCTURE UNDER MOVING MASS
USING INTELLIGENT METHODS
Thesis Submitted to the
Department of Mechanical Engineering
National Institute of Technology, Rourkela
in partial fulfilment of the requirements
of the degree of
Doctor of Philosophy
in
Mechanical Engineering
by
Shakti Prasanna Jena
(Roll Number: 512ME123)
under the supervision of
Prof. Dayal R. Parhi
December 2016
Department of Mechanical Engineering
National Institute of Technology Rourkela
i
Mechanical Engineering
National Institute of Technology, Rourkela
December 7, 2016
Certificate of Examination
Roll Number: 512ME123
Name: Shakti Prasanna Jena
Title of Dissertation: Dynamic Analysis and Fault Detection of Multi Cracked Structure
under Moving Mass using Intelligent Methods
We, the below signed, after checking the dissertation mentioned above and the official
record book (s) of the student, hereby state our approval of the dissertation submitted in
partial fulfilment of the requirements of the degree of Doctor of Philosophy in Mechanical
Engineering at National Institute of Technology Rourkela. We are satisfied with the
volume, quality, correctness, and originality of the work.
---------------------------
Dayal R. Parhi
Principal Supervisor
--------------------------- ----- -----------------------------
S. Murugan R. Mazumder
Member (DSC) Examiner Member (DSC)
------------------------------- ------------------------------
S. K. Das K V Sai Srinadh
Member (DSC) Examiner External Examiner
------------------------------ ------------------------------
K. P. Maity S. S. Mahapatra
DSC Chairman Head of the Department
ii
Mechanical Engineering
National Institute of Technology, Rourkela
Supervisor Certificate
This is to certify that the work presented in this thesis entitled “Dynamic Analysis and
Fault Detection of Multi Cracked Structure under Moving Mass using Intelligent
Methods” by “Shakti Prasanna Jena”, Roll Number 512ME123, is a record of original
research carried out by him under my supervision and guidance for the partial fulfilment
of the requirements of the degree of Doctor of Philosophy in the Department of
Mechanical Engineering, National Institute of Technology, Rourkela, Odisha, India.
To the best of my knowledge, neither this thesis nor any part of it has been submitted for
any degree or diploma to any institute or university in India or abroad.
--------------------------
December 7, 2016 Supervisor
(Dr Dayal R. Parhi )
Professor
Department of Mechanical Engineering
National Institute of Technology, Rourkela
Odisha, India
iii
Declaration of Originality
I, Shakti Prasanna Jena, Roll Number 512ME123 hereby declare that this dissertation
entitled “Dynamic Analysis and Fault Detection of Multi Cracked Structure under
Moving Mass using Intelligent Methods” represents my original work carried out as a
doctoral student of NIT, Rourkela and, to the best of my knowledge, it contains no
material previously published or written by another person, nor any material presented for
the award of any other degree or diploma of NIT, Rourkela or any other institution. Any
contribution made to this research by others, with whom I have worked at NIT, Rourkela
or elsewhere, is explicitly acknowledged in the dissertation. Works of other authors cited
in this dissertation have been duly acknowledged under the section ''Bibliography''. I have
also submitted my original research records to the scrutiny committee for evaluation of my
dissertation. I am fully aware that in case of any non-compliance detected in future, the
Senate of NIT, Rourkela may withdraw the degree awarded to me on the basis of the
present dissertation.
December 7, 2016 Shakti Prasanna Jena
NIT Rourkela
iv
Acknowledgements
Though only my name appears on the cover of this dissertation, many people have contributed to
its production. I owe my gratitude to all those people who have made this dissertation possible and
because of whom my doctorate experience has been one that I will cherish forever.
My deepest gratitude is to my supervisor, Prof. Dayal R. Parhi. I have been amazingly fortunate to
have an advisor who gave me the freedom to explore on my own and at the same time the
guidance to recover when my steps faltered. His patience and support helped me overcome many
crisis situations and finish this dissertation. I hope that one day I would become as good an advisor
to my students as he has been to me.
I am thankful to Prof. Animesh Biswas, Director of National Institute of Technology, for giving
me an opportunity to be a part of this institute of national importance and to work under the
supervision of Prof. Dayal R. Parhi. I am thankful to Prof. S. S. Mahapatra, Head of the
Department, Department of Mechanical Engineering, for his moral support and valuable
suggestions during the research work.
I express my gratitude to Prof. K.P. Maity (Chairman DSC) and DSC members for their indebted
help and valuable suggestions for accomplishment of dissertation.
I thank all the members of the Department of Mechanical Engineering, and the Institute,
who helped me in various ways towards the completion of my work.
I would like to thank all my friends (Durga, Madhu, Sidha and Smruti) and lab-mates
(Prases, Biplab, Kaku, Animesh, Subhasri, Sasmita, Anish, Yadoo, Chinu, Ghodki, Alok, Prabir
and Irshad) in Robotics lab for their encouragement and understanding. Their support and lots of
lovely memory with them can never be captured in words. My special thanks to Mr Prases
Mohanty helping me a lot and working in the laboratory for late nights. I want to especially
thankful Mr. Mahesh, for helping me in various ways throughout my Ph.D work.
I thank my parents, my grandmother, my uncles (Lalit & Bipin) and entire family members for
their unlimited support and strength. Without their dedication and dependability, I could not
have pursued my PhD degree at the NIT, Rourkela. I would like to give special thanks to my wife
(Reena) for her continuous supports and encouragements in my days of PhD work.
Last, but not the least, I praise the Almighty for giving me the strength during the research work.
December 7, 2016 Shakti Prasanna Jena
NIT Rourkela Roll Number: 512ME123
v
Abstract
The present thesis explores an inclusive research in the era of moving load dynamic
problems. The responses of vibrating structures due to the moving object and different
methodologies for damaged identification process have been investigated in this analogy.
The theoretical-numerical solutions of the multi-cracked structure with different end
conditions subjected to transit mass have been formulated. The Runge-Kutta fourth order
integration approach has been applied to determine the response of the structures
numerically. The effects of parameters like mass and speed of the traversing object, crack
locations, and depth on the response of the structures are investigated. The proposed
numerical method has been verified using FEA and experimental investigations. The novel
damage prediction processes are developed on the knowledge-based concepts of recurrent
neural networks (RNNs) and statistical process control (SPC) methods as inverse
approaches. The Jordan’s recurrent neural networks (JRNNs), Elman’s recurrent neural
network (ERNNs), the integrated approach of the JRNNs, and ERNNs, the autoregressive
(AR) process in the domain of SPC and the combined hybrid neuro-autoregressive process
have been developed to identify and quantify the faults in the structure. The accuracy and
exactness of each approach has been verified with experiments and FEA. The proposed
methods can be useful for the online condition monitoring of faulty cracks in structures.
Keywords: cracked beam, Runge-Kutta, RNNs and autoregressive.
vi
Content
Certificate of Examination ................................................................................................. i
Supervisor Certificate ....................................................................................................... ii
Declaration of Originality ................................................................................................ iii
Acknowledgement ............................................................................................................. iv
Abstract .............................................................................................................................. v
Contents ............................................................................................................................. vi
List of Figures ................................................................................................................... ix
List of Tables .................................................................................................................... xii
Nomenclatures ................................................................................................................ xiv
1 Introduction ................................................................................................................. 1
1.1 Motivation ............................................................................................................... 1
1.2 Objective of the Research Works ............................................................................ 2
1.3 Novelty of the Research Work ................................................................................ 3
1.4 Contributions ........................................................................................................... 4
1.5 Overview of the thesis ............................................................................................. 4
2 Literature Review ........................................................................................................ 6
2.1 Introduction ............................................................................................................. 6
2.2 Response analysis of structure ................................................................................. 6
2.2.1 Classical methods ........................................................................................... 6
2.2.2 Finite element analysis/method for response analysis of structure ............... 12
2.3 Damage detection .................................................................................................. 15
2.3.1 Classical/FEA based Methods for damage detection in structure ................ 15
2.3.2 AI techniques based methods for fault detection in structure ....................... 22
2.3.2.1 Genetic Algorithm based methods for crack detection in structure . 22
2.3.2.2 Neural network based methods for crack detection in structure ...... 23
2.3.2.3 Recurrent neural network based methods for crack detection in
structure ........................................................................................................ 27
2.4 Statistical based method for fault identification in structure ................................. 28
2.5 Miscellaneous methods .......................................................................................... 29
2.6 Summary ................................................................................................................ 30
vii
3 Theoretical-Numerical Analysis of Multi-cracked Structures Subjected to
Moving Mass .............................................................................................................. 32
3.1 Introduction ........................................................................................................... 32
3.2 The problem Description ....................................................................................... 32
3.3 The Problem Formulation ...................................................................................... 33
3.4 Analysis of Cracked Structures Subjected to Moving Mass ................................. 33
3.5 Numerical Formulation of Cracked Cantilever Beam under a Moving Mass ....... 37
3.6 Theoretical-Numerical Solution of Cracked Simply Supported Structure under a
Moving Mass ......................................................................................................... 45
3.7 Theoretical-Numerical Solution of Cracked Fixed-Fixed Beam under a Moving
Mass ....................................................................................................................... 53
3.8 Identification of cracks from the measured dynamic response of structures ........ 61
3.9 Comparison of Results of Theoretical-Numerical Experimental analysis for the
Response of Structures ........................................................................................... 64
3.10 Discussions and Summary ................................................................................... 67
4 Finite Element Analysis of Cracked Structures Subjected to Moving Mass ....... 70
4.1 Introduction ........................................................................................................... 70
4.2 Method for FEA of moving mass-structure using ANSYS ................................... 70
4.3 Steps involving 'The full method' transient dynamic analysis in ANSYS ............ 72
4.4 Response analysis of cracked structures under moving mass using ANSYS ........ 72
4.5 Discussion and Summary ...................................................................................... 84
5 Application of Recurrent Neural Networks for Damage Identification in
Structures Under Moving Mass ............................................................................... 86
5.1 Introduction ........................................................................................................... 86
5.2 Overview of neural networks ................................................................................ 87
5.2.1 Feed forward neural networks ...................................................................... 87
5.2.2 Recurrent neural networks (RNNs) .............................................................. 88
5.3 Use of Levenberg-Marquardt back propagation method for RNN ........................ 92
5.3.1 Steps for the organization of the training procedure using L.M algorithm .. 93
5.4 Application of rule-based modified JRNNs for damage identification in
structure under moving mass ................................................................................. 94
5.5 Application of rule-based modified FRNNs for damage detection in structure
subjected to moving mass ...................................................................................... 98
5.6 Application of rule-based modified hybridized JRNNs and ERNNs for
viii
for multiple damage detection in structure subjected to moving mass ................. 101
5.7 Discussion and Summary .................................................................................... 113
6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass .............................................................................. 114
6.1 Introduction ......................................................................................................... 114
6.2 Overview of Statistical Process Control (SPC) method ...................................... 115
6.3 Construction and analysis of control chart .......................................................... 115
6.4 Overview of Autoregressive model ..................................................................... 117
6.5 Application of auto regressive (AR) model based method for damage detection
in structures subjected to traversing mass ............................................................ 119
6.6 Discussion and Summary .................................................................................... 130
7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection in Beam
Structures Subjected to Moving Mass ................................................................... 132
7.1 Introduction ......................................................................................................... 132
7.2 Development of combined hybrid neuro-autoregressive model for damage
detection in beam type structures subjected to moving mass ............................... 133
7.3 Discussion and Summary .................................................................................... 138
8 Experimental Analysis of Damaged Structures Subjected to Transit Mass ...... 139
8.1 Introduction ........................................................................................................ 139
8.2 Experimental Procedure ...................................................................................... 139
8.3 Discussion and Summary .................................................................................... 143
9 Results and Discussion ............................................................................................ 145
9.1 Introduction ........................................................................................................ 145
9.2 Analysis and results of different adopted methods .............................................. 145
9.3 Summary .............................................................................................................. 149
10 Conclusions and Suggestion for Further Research .............................................. 150
10.1 Introduction ...................................................................................................... 150
10.2 Contributions ..................................................................................................... 150
10.3 Conclusions ....................................................................................................... 151
10.4 Recommendation for future study ..................................................................... 152
Appendix ......................................................................................................................... 153
Bibliography ................................................................................................................... 156
Dissemination ................................................................................................................. 176
Vitae ............................................................................................................................. 178
ix
List of Figures
3.1 Multi- cracked cantilever beam under moving mass .......................................... 34
3.2 For undamaged beam for 438 /v cm s ............................................................... 38
3.3 For undamaged beam for 573 /v cm s ............................................................... 38
3.4 For 1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s ............................... 39
3.5 For 1,2,3 1,2,3573 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s ................................ 39
3.6 For 1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.5,0.65,0.85v cm s ................................. 40
3.7 For 1,2,3 1,2,3573 / , 0.6,0.25,0.45. 0.5,0.65,0.85v cm s ................................. 40
3.8 For 1,2,3 1,2,3438 / , 0.3,0.55,0.4. 0.25,0.45,0.65v cm s ................................. 41
3.9 For 1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.25,0.45,0.65v cm s .................................. 41
3.10 For 1,2,3 1,2,3438 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s ................................. 42
3.11 For 1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s ................................. 42
3.12 3-D Graph for time mass deflection for
1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s ........................................ 43
3.13 3-D Graph for time mass deflection for
1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s ………………………………….43
3.14 3-D Graph for time speed deflection for
1,2,3 1,2,32 , 0.6,0.25,0.45. 0.25,0.45,0.65M kg ........................................... 44
3.15 3-D Graph for position time deflection for
1,2,3 1,2,32 , 573 / , 0.6,0.25,0.45. 0.25,0.45,0.65M kg v cm s ....................... 44
3.16 Multi-cracked simply supported beam under moving mass .............................. 45
3.17 For undamaged beam for 438 /v cm s ............................................................. 46
3.18 For undamaged beam for 573 /v m s ............................................................... 46
3.19 For 1,2,3 1,2,3438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s ........................... 47
3.20 For 1,2,3 1,2,3573 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s ........................... 47
3.21 For 1,2,3 1,2,3438 / , 0.35,0.45,0.55. 0.2857,0.5,0.7143.v cm s ..................... 48
3.22 For 1,2,3 1,2,3573 / , 0.35,0.45,0.55. 0.2857,0.5,0.7143.v cm s ..................... 48
3.23 For 1,2,3 1,2,3438 / , 0.2,0.3,0.4. 0.1786,0.3571,0.5714.v cm s ...................... 49
3.24 For 1,2,3 1,2,3573 / , 0.2,0.3,0.4. 0.1786,0.3571,0.5714.v cm s …….. ........... 49
3.25 For 1,2,3 1,2,3438 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s ................ 50
3.26 For 1,2,3 1,2,3573 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s ............ …50
3.27 3-D Graph for time ~mass ~deflection for
x
1,2,3 1,2,3 573 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714v cm s .... ……………51
3.28 3-D Graph for time speed deflection for
1,2,3 1,2,3 2 , 0.35,0.45,0.55. 0.1786,0.3571,0.5714M kg …………………………...51
3.29 3-D Graph for time mass deflection for
1,2,3 1,2,3 438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143v cm s ................................... 52
3.30 3-D Graph for position ~ time ~ deflection for
1,2,3 1,2,3 1 , 438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143M kg v cm s .................. …52
3.31 Multi-cracked fixed-fixed beam subjected to moving mass .............................. 53
3.32 For undamaged beam for 512 /v cm s …………………………………….......54
3.33 For undamaged beam for ................................................................................... 54
3.34 For 1,2,3 1,2,3512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357v cm s ........... …….55
3.35 For 1,2,3 1,2,3617 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357v cm s .................... 55
3.36 For 1,2,3 1,2,3512 / , 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s ....................... 56
3.37 For 1,2,3 1,2,3617 / , 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s ....................... 56
3.38 For 1,2,3 1,2,3512 / , 0.3,0.5,0.55. 0.1429,0.3214,0.5357v cm s ...................... 57
3.39 For 1,2,3 1,2,3617 / , 0.3,0.5,0.55. 0.1429,0.3214,0.5357v cm s ...................... 57
3.40 For 1,2,3 1,2,3512 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143v cm s .......................... 58
3.41 For 1,2,3 1,2,3617 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143v cm s .......................... 58
3.42 3-D Graph for time mass deflection for
1,2,3 1,2,3 617 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143.v cm s .............................. 59
3.43 3-D Graph for time speed deflection for
1,2,3 1,2,3 2 , 0.3,0.5,0.55. 0.25,0.4286,0.7143.M kg ...................................... 59
3.44 3-D Graph for time mass deflection for
1,2,3 1,2,3 512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357.v cm s ......................... 60
3.45 3-D Graph for Position time deflection for 2 ,M kg
1,2,3 1,2,3 512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357.v cm s ......................... 60
3.46 Detection of cracks for cantilever beam for 1,2,3 0.5,0.65,0.85 ...................... 61
3.47 Detection of cracks for simply supported beam for
1,2,3 0.2857,0.5,0.7143. ................................................................................ ....62
3.48 Detection of cracks for fixed-fixed beam for 1,2,3 0.25,0.4286,0.7143. …..….62
3.49a Magnified view of crack for 0.5 ………………………...…………………63
3.49b Magnified view of crack for 0.65 ………………………...…………………63
3.49c Magnified view of crack for 0.85 ..………………………….…………….64
4.1 Free body diagram of vibrating system……………………………………...…71
xi
4.2 Transit mass-structure interaction of cracked cantilever beam for
1,2,3 1,2,30.3,0.55,0.4. 0.5,0.65,0.85. M=2 kg…………..…………………....73
4.3 Magnified view of crack for α=0.55 ......................................................... …….73
4.4a Second mode shape of cantilever structure .............................................. …….74
4.4b Third mode shape of cantilever structure ................................................. …….75
4.5 Schematic view of transient structural model for cracked cantilever beam…...75
4.6 For cracked cantilever beam for 1,2,3 1,2,30.3,0.55,0.4. 0.25,0.65,0.85 ....... 77
4.7a Second mode shape of simply supported beam ................................................. 78
4.7b Third mode shape of simply supported beam .................................................... 79
4.8 For cracked simply supported beam for, 2 , 438 /M kg v cm s
1,2,3 1,2,30.35,0.45,0.55. 0.2857,0.5,0.7143 .................................................. 81
4.9a Second mode shape fixed-fixed beam ............................................................... 81
4.9b Third mode shape fixed-fixed beam .................................................................. 82
4.10 For cracked fixed-fixed beam for 2 , 617 /M kg v cm s ,
1,2,3 1,2,30.2,0.35,0.45. 0.1429,0.3214,0.5357. ............................................ 83
5.1 Simplified NN model with feed forward networks ................................... …….87
5.2 Architecture of a RNN model ............................................................................. 88
5.3 Simple Architecture of JRNN model ................................................................. 90
5.4 Simple architecture of ERNN model .............................................................. …90
5.5 Simple architecture of HRNN model ........................................................ …….91
5.6 Architecture of modified JRNN model ..................................................... …….95
5.7 Architecture of modified ERNN model .................................................... …….99
5.8 Hybridized architecture of modified JRNN and ERNN models ...................... 102
5.9 Plot of graph of iterations vs. sum square errors for RNNs methods model .... 104
6.1 Architecture of control chart ................................................................... …….116
6.2 Representation of a probability process as the outputs from a liner filter…….117
6.3 Data analysis in SPSS windows ....................................................................... 120
6.4a Control chart for cantilever beam .................................................................... 120
6.4b Control chart for simply supported beam ........................................................ 121
6.4c Control chart for fixed-fixed beam .................................................................. 121
7.1 Architecture of Hybrid neuro autoregressive model ....................................... 133
8.1 Experimental set up for cantilever beam ..................................................... …140
8.2 Experimental set up for simply supported beam .................................... …….141
8.3 Experimental set up for fixed-fixed beam .............................................. …….141
8.4 Ultrasonic sensor ..................................................................................... …….142
8.5 Micro-controller Aurdino ............................................................................... 142
8.6 Damaged portion of beam .............................................................................. 143
8.7 Variac........................................................................................................144
8.8 Bread board...............................................................................................144
xii
9.1 Comparison of results among different damage detection methods .............. 149
A1 Crack analysis on the vibration characteristics of cantilever beam .............. …153
List of Tables
3.1 Comparison of results between experiment and numerical for cracked cantilever
beam for 1,2,3 1,2,3438 / . 0.6,0.25,0.45. 0.25,0.45,0.65.v cm s ...................... 65
3.2 Comparison of results between experiment and numerical for cracked simply
supported beam for 1,2,3 1,2,3573 / . 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s .... 66
3.3 Comparison of results between experiment and numerical for cracked
fixed-fixed beam for 1,2,3 1,2,3617 / . 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s ..66
4.1 Frequencies ratios of damaged cantilever beam .................................................. 74
4.2 Comparison of results between experiment and FEA for cracked cantilever beam
for 1,2,3 1,2,3573 / . 0.3,0.55,0.4. 0.5,0.65,0.85.v cm s .................................... 76
4.3 Comparison of results among experiment, FEA and numerical for cracked
cantilever beam for M=1kg,v=438cm/s, 1,2,3 1,2,30.3,0.55,0.4. 0.5,0.65,0.85 ..... 77
4.4 Frequencies ratios of damaged simply supported beam ...................................... 79
4.5 Comparison of results between experiment and FEA for cracked simply
supported beam for 1,2,3 1,2,3438 / . 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s .... 79
4.6 Comparison of results among experiment, FEA and numerical for cracked simply
supported beam for M=1kg, v=573cm/s, 1,2,3 1,2,30.35,0.45,0.55. 0.1876,0.3571,0.5714. ...80
4.7 Frequencies ratios of damaged fixed-fixed beam…………………………….…80
4.8 Comparison of results between experiment and FEA for cracked fixed-fixed
beam for 1,2,3 1,2,3512 / . 0.3,0.5,0.55. 0.25,0.4286,0.7143.v cm s ................. 82
4.9 Comparison of results among experiment, FEA and numerical for cracked
fixed-fixed beam for M=2kg, v=617cm/s, 1,2,3 1,2,30.3,0.5,0.55. 0.25,0.4286,0.7143. ..83
5.1 Test patterns to the RNN model for cracked cantilever beam ........................... 104
5.2(a, b) Comparison of results between experiments and different RNNs method
for prediction of relative crack locations (cantilever beam)………………105-106
5.3(a,b) Comparison of results between experiments and different RNNs method
for prediction of relative crack locations (simply supported beam) ............ 107-108
5.4(a,b) Comparison of results between experiments and different RNNs method
for prediction of relative crack locations (fixed-fixed beam) ....................... 109-110
5.5(a,b) Comparison of results among various methods for prediction of relative
crack locations (fixed-fixed beam) ............................................................. 111-112
xiii
6.1(a,b) Comparison of results among FEA, Theoretical and SPC methods for
estimation of relative crack locations (cantilever beam) ............................ 125-126
6.2(a,b) Comparison of results among FEA, Theoretical and SPC methods for
estimation of relative crack locations (simply supported beam) ......... 126-127
6.3(a,b) Comparison of results among Experimental, FEA, Theoretical and
SPC methods for estimation of relative crack locations (fixed-fixed
beam)…………………………………………………………………128-129
7.1(a,b) Comparison of results among FEA, Theoretical and hybrid approach
of RNNs, and SPC methods for prediction of relative crack locations
(cantilever beam) ........................................................................................ 135
7.2(a,b) Comparison of results among Experimental, FEA, Theoretical and
hybrid approach of RNNs, and SPC methods for prediction of relative
crack locations (simply supported beam) ................................................... 136
7.3(a,b) Comparison of results among FEA, Theoretical and hybrid approach
of RNNs, and SPC methods for prediction of relative crack locations
(fixed-fixed beam) ....................................................................................... 137
8.1 Specification of different components of experimental set up .......................... 143
xiv
Nomenclature
( )u x u = Deflection due to longitudinal vibration of the beam. ( )y x y = Deflection due to transverse vibration of the beam. ( )I x I = Moment of inertia of the beam.
( )m x m = Beam mass per unit span length.
g = gravitational constant
L =Length of the beam.
( )F t The interactive force between the moving mass and the structure.
1 2 3 1,2,3, ,L L L L = Position of the first, second and third cracks at the left end of the
cracked structures respectively.
1,2,31,2,3
L
L =Relative crack Positions from the fixed end.
1 2 3 1,2,3, ,d d d d = Depth of the first, second and third cracks respectively.
1,2,31,2,3
dH
= Relative crack depth
M = Mass of the moving mass
v =Moving speed of the transit mass.
vt = Position of the moving mass at any instant time‘t’.
h = The point of attention where the beam deflection is to be determined.
( , )r x t Standard loading conditions.
ϐ = Sheer force, = Bending moment.
δ=Dirac delta function.
( )n x = Eigen functions of the beam.
( )nQ t = Amplitude functions of the beam.
n = Number modes of vibrations.
[ ]tM x -Inertial force
[ ]tC x -Damping force
[ ]tK x -Stiffness force,
xv
( )f t -Applied force.
ρ- Density of the beam.
E- Young’s modulus of elasticity.
υ- Poisson’s ratio
k- Stiffness
G – Rigidity of the beam.
f(.) and g(.) - Activation functions in the hidden and output layers respectively.
w- Synaptic weights.
W-Input values to the neural networks.
- Output values from the neural networks.
1Z - The delay units of the neural networks.
J - The Jacobian matrix.
ξ- The combination coefficient.
ν - The step size or training constant.
X - Identity matrix.
e - The error value.
- The error function.
- The self-recurrent values of the each node.
- The learning rate of the neural network.
- The mean. w - The standard deviation.
( )b - Autoregressive operator.
s - The sample size.
ta - Shock or white noise, c- Order of the autoregressive process
- Co-efficient of autoregressive process.
- The covariance matrix.
V - The variance.
-The probability density function.
s - The statistical moment.
tz - The linear filter.
1
Chapter 1
INTRODUCTION
For most part, structures in civil, mechanical and aerospace engineering applications are
subjected to space and time varying masses or loads. Moving masses or loads have
significant consequence on the dynamic characteristics of the structures. Dynamic analysis
of structures under the action of moving mass is the traditional subject of research. The
problems of the dynamic response of structures subjected to moving mass have been
investigated by numerous scientists, engineers, researchers and mathematicians for last
few decades. But the first incident of moving mass problem came in 1847; the collapse of
the Chester Bridge, England [Atkin and Mofid, 3], which eventually led heartbreaking
hammering of human beings. The solutions of the moving mass problems draw attentions
of several scientists, engineers, researchers and mathematicians to investigate the moving
mass problem. Initially, the solutions for the moving mass problems are formulated on the
basic assumption, and more often, ignoring the mass of the traversing object. Later, the
problem involves the inclusions of inertial, rotary, shearing and damping effects. But due
to the existence of different types of cracks, increased mass and speed of the moving
object, and structural boundary condition, the problems become more complicated.
Therefore, enhanced learning mechanism is necessary to investigate the response of
structures, early detection of crack locations and severities to improve the service periods
of structure, mass, human beings and better economic growth.
This Chapter describes the background and motivation, objectives and scopes, and
overviews of the present thesis.
1.1 Motivation
The applications of engineering structures under moving mass have gained enormous
importance in the field of aerospace, buildings, defence, cranes, bridges, railways
engineering, etc. The dynamic behaviour of moving mass-structural systems has been
characterised by geometrical and mathematical modelling along with the consideration of
initial conditions, different end conditions, geometric characters, material properties, basic
assumptions and a set of coupled partial differential equations. The occurrence of faults in
Chapter 1 Introduction
2
engineering structures may be unavoidable in spite of our greatest efforts. The faults may
come concerning cracks, fatigue, enhanced moving speed of the mass, enhancement of
mass, mechanised progression, erosion, etc. The faults are induced in the structure due to
several reasons such as metallurgical defects, manufacturing defects, environmental
defects and mechanical defects. So it is important to study the integrity and consistency of
structures for better service periods.
Various analyses, experiments and methodologies developed by engineers, scientists and
researchers, provide the inspiration and the support for the innovative research. The
existence of faults deteriorates the behaviour of the structures up to a certain degree. So it
is required for the early detection of faults for sustaining the stability and existence of
structures. There are several methods developed by various researchers for fault diagnosis
of structures under moving mass. Most of the methods (gamma radiation, ultrasonic
testing, X-ray testing, magnetic particle and etc.) are limited to local damage isolation, less
sensitive and costly. But vibration based damage isolation methods are global and highly
sensitive. So Artificial Intelligence (AI) techniques as damage detector are developed by
changing the dynamic properties and vibration signatures of structures by many
researchers which are global and cost-efficient. Here some rule-based Recurrent Neural
Network methods, statistics based method along with some novel hybridised methods are
proposed to identify, estimate and quantify the extent and severities of cracks. The
responses of different types of structures under traversing mass with multiple cracks are
also determined using analytical- numerical solutions along with FEA and experimental
verifications.
1.2 Objectives of the Research Works
The effect of moving load over a structure is the fundamental problem in structural
dynamics. The studies of response and the parameters influencing the responses of
structures to moving mass are essential for the integrity and health monitoring of
structures. The primary objectivities of the current thesis are to determine the response of
the multi-cracked structure under moving mass with different boundary states using
theoretical, FEA and experimental analyses, and to develop reverse methodology on crack
detection. The specific objectives of the present thesis are:
Chapter 1 Introduction
3
1) The presentation of simple and sensible methods for determining the response of multi-
cracked structures with different end conditions subjected to a moving mass.
2) The formulation of solution to the problem in analytical-numerical form.
3) The verification to the solution of the problem obtained from the analytical-numerical
method using FEA and experimental analysis.
4) The analyses of the significance of various parameters and its influences on the
response of the structures.
5) The development of damage isolation procedure based on AI technique using the
change in dynamic response of the structure.
6) The development of rule-based RNNs (Jordan and Elman) for damage isolation.
7) The hybridisation of Jordan’s and Elman’s RNNs for damage isolation.
8) The development of a novel damage isolation method based on statistical data analysis.
9) The development of a novel hybridised method using the rule-based RNNs and
statistical data analysis for damage identification.
1.3 Novelty of the Research Work
The present research work is concerned with the determination of the response of
structures under the influence of moving mass and the development of novel damage
isolation procedures based on improved vibration-based damage identification method.
The novelty of the proposed work is restricted to the subsequent areas:
1) The investigation of the theoretical-numerical method along with FEA and
experimental verifications.
2) The consideration of different parameters and their influences on the response of the
structures subjected to moving mass.
3) The investigation of novel AI techniques for damage isolation in structures based on the
concepts of RNNs (Jordan’s, Elman’s and hybridisation of Jordan’s, and Elman’s RNNs),
statistical process control method and vibration induced data.
To the best of Author’s knowledge, the applications of knowledge-based RNNs methods
for damage detection in structures are not reported previously. The damage detection
method using the displacement response of structure in statistical process control analysis
is not cited earlier. The application of the combined hybrid neuro-autoregressive process
for structural damage detection is scanty.
Chapter 1 Introduction
4
1.4 Contributions
The most important role of this research work is the improvement and validation of novel
vibration-based damage isolation methods. The vibration-based damage isolation methods
include the formulation of rule-based RNNs (Jordan and Elman) method. This method is
also based on statistical data analysis and the integration of the hybridised RNNs, and
statistical data analysis method. The proposed methods have been applied to improve the
original damage isolation methods for better prediction of the damage identification,
localisation and quantification of damage severities. All the methods have been verified
with numerical, FEA and experimental tests.
1.5 Overview of the thesis
The entire thesis includes ten Chapters. Brief descriptions of all the Chapters are as
follows:
Chapter-1 presents the introductory knowledge on moving load problems. This Chapter
includes the motivation behind this study, research aim and objectives, the scope of the
research and summarisation of the present work.
Chapter-2 explains about research works carried out by various researchers. This Chapter
presents various methodologies, analyses and experiments on moving load problem. This
Chapter gives the insufficiency in literature reviews and formulates the novelty of the
present study.
Chapter-3 formulates an analytical-numerical method for the solution of different types
of structures under moving mass.
Chapter-4 carries out FEA for the solution of the moving mass problem on structure.
Chapter-5 formulates the fault analysis of multiple cracked structures subjected to
moving mass using different types of rule-based RNNs.
Chapter-6 explains the formulation of statistics based damage isolation method for the
multi-cracked structure under moving mass.
Chapter-7 presents the integrated approach of the hybridised RNN method and the rule-
based statistics based method.
Chapter-8 conducts laboratory tests for different types of structures under moving mass.
Chapter-9 explains about the reviews of all the methodologies made in the thesis and
shows the comparison of results.
Chapter 1 Introduction
5
Chapter-10 discusses conclusions and scope of future work to be carried out.
6
Chapter 2
Literature Review
2.1 Introduction
This Chapter discusses the innovative works reported by various researchers in the field of
moving load structure interaction dynamics and its applications in various engineering,
transportation and infrastructure industries. Numerous techniques have been proposed to
determine the response and to diagnose faults in the structure under moving mass. Several
significant works of researchers which have received great attentions are explained briefly
in this Chapter.
2.2 Response analysis of structure
2.2.1 Classical methods
Sridharan and Mallik [1] carried out a numerical analysis to study the vibrational effects
beams under moving load. They employed the Wilson-θ method to analyze the response
of the structure. Siddiqui et al. [2] have investigated the dynamic response of a cracked
cantilever structure under a transit mass with internal resonance conditions. The response
of the moving mass-structure has been determined using the Rayleigh-Ritz method and the
perturbation method of multiple scales to obtain approximate solution. A theoretical-
numerical method has been developed by Akin and Mofid [3] to carry out the response of
the structure under moving mass using Runge-Kutta method. They have compared the
results from numerical analysis with finite element analysis (FEA) and found good
agreements. Stanisic and Hardin [4] have developed a theory to describe the response of
the structure under a random number of traversing masses using Fourier’s analysis.
Olsson [5] discussed the dynamic analysis of a simply supported structure under a
constant moving force travelling at constant speed. He has presented analytical and finite
element solution to this simply supported beam problem. Mofid and Akin [6] have
presented a novel experimental and an inexact method to calculate the response of the
structure under a moving load with respect to time. This inexact method is applicable for
wide range of structure with various end conditions. Kwon et al. [7] examined the analysis
Chapter 2 Literature Review
7
of vibration and control of bridge structure subjected to moving load using Tuned Mass
Damper (TMD) method. Mahmoud and Abouzaid [8] presented an iterative modal
analysis technique to examine the consequence of transverse crack on the response of a
simply supported structure subjected to travelling mass. They have explained the exact
effects of cracks and mass depending on the various parameters like traversing time,
moving speed, the location of cracks and crack types, etc.
Li et al. [9] have extracted the natural frequencies of railway girder bridges subjected to
moving load by using the eigenvalues solution of the moving load-bridge interaction
dynamics. Bilello et al. [10] conducted experimental verifications of a simply supported
highway bridge under a moving vehicle. They have applied the theory of structural model
to enlarge scale model experiment and to study the response of the bridge structure.
Bilello and Bergman [11] performed a theoretical investigation along with experimental
verification for the analysis of a cracked beam under travelling mass. Yang et al. [12]
presented a study to extort the fundamental frequency of a bridge structure by means of
the dynamic response of the moving vehicle across the bridge. Majka and Hartnett [13]
developed a proficient numerical model to investigate the consequence of different
parameters affecting the response of railway bridge structure. It has been observed that
parameters like the ratio of the train to bridge frequency, the speed of the train, span ratio,
mass and bridge damping are influencing the response of the structure.
Garinei and Risitano [14] have conducted a brief analysis of the conditions occurring from
the mutual loads spread over a single axle and mutual loads spread over a series of
equidistant loads of a railway bridge structure under moving load. Yang and Chang [15]
have conducted parametric studies to extract the bridge frequency indirectly from a
passing vehicle. Dehestani et al. [16] have developed a theoretical analysis followed by
computational analysis to study the response structure under moving mass with various
boundary conditions. The critical influential speed has been introduced in the analysis and
calculated numerically for different types of structures at various boundary conditions. Liu
et al. [17] have investigated the conditions of train-bridge interaction dynamics to study
the vibration induced due to the train-bridge interaction. The consequences of the train-
bridge interaction dynamics on the bridge response has been investigated at the time of the
passage of the train. The influence of different parameters like critical speed, the
frequency ratio of vehicle to bridge, vehicle model types and high modal damping value
have been investigated on the response of the bridge during the passage of the vehicle.
Chapter 2 Literature Review
8
Siringoringo and Fujino [18] have carried out an analytical and experimental study to
approximate the bridge fundamental frequency indirectly from the dynamic response of
traversing vehicle. The vehicle load with pulsation sensor has been used to carry out
periodic measurement over various bridge structures and approximated their fundamental
frequencies. Xia et al. [19] investigated an analytical expression along with numerical
simulation and experimental verifications to investigate the resonance mechanism and
conditions of train-bridge structure. They observed that the resonance of the vehicle was
induced due to the periodical arrangement of bridge span and deflection. Majkaa and
Hartnett [20] have carried out a study to examine the dynamic effects induced by service
train, the significance of track randomness and bridge skewness. The extent of the
research reached at the observation that the significance of random track irregularities had
a small impact on the dynamic magnification factors and bridge acceleration, where as
track irregularities had the greater impact on bridge response.
A theoretical and experimental solution are carried out to explore the dynamic behaviour
of ground vibration induced due to a high-speed railway train, train on bridges and trains
in tunnels by Ju et al. [21]. The results indicated that the ground vibration induced due to
the train at the train load dominant frequencies are considerably great for both subsonic
and supersonic speeds of the train. Yoon et al. [22] have formulated a theoretical and
experimental analysis to examine the free vibration characteristics of a double cracked
simple supported Euler-Bernoulli beam with open cracks. The consequences of crack
depth and location on the natural frequency of the simple cracked beam were examined.
Mahmoud [23] has presented a method to calculate the stress intensity factor of a cracked
beam under travelling load with single and double edge cracks. Modal analysis method
was applied to calculate the equivalent load on the cracked structure and stress intensity
factor was calculated using the concepts of linear fracture mechanics. It has been observed
that stress intensity factor depends on moving mass speed, time, location and size of the
crack. Michaltsos et al. [24] have discussed the significance of traversing load and other
constraints on the response of a simply supported structure under moving mass. Mahmoud
[25] has described the response of a cracked, undamped simply supported beam under
traversing mass in the presence of a transverse crack. The effects of cracks on the response
of the simple beam were investigated through numerical solution.
Lin and Chang [26] obtained the response of a damaged cantilever beam along a
concentrated traversing load using the theoretical transfer matrix method. In the later part,
Chapter 2 Literature Review
9
the forced response of the damaged structure was found out by using modal expansion
theory using the calculated eigenfunctions. Ouyang [27] carried out a tutorial on moving
load dynamic problems. Numerous types of essential concept related to structural dynamic
problems are explained in this study. Using the discrete element technique and finite
element method, Ariaei et al. [28] proposed a theoretical as well as calculation method to
obtain the response of a damaged structure under moving mass with open and breathing
cracks. The consequence of crack on the resonance condition of the structure was also
inspected. It was observed that crack can alter the critical speed leading to the resonance
of the structure. Azam et al. [29] have studied the response of a Timoshenko beam under
traversing and sprung mass structures. From various numerical examples, it was seen that
the deflection obtained due to moving mass system are higher than that of moving sprung
systems.
Michaltsos and Kounadis [30] have examined the consequence of centripetal, and Coriolis
forces on the response of a light girder bridge subjected to traversing load. Pala and Reis
[31-32] have investigated the significance of inertial, Coriolis, and centripetal, forces on
the dynamic response of cracked structures subjected to a traversing load with a single
crack. The response of the cracked structure was calculated using Duhamel integral. It was
concluded that the mass and speed of the traversing load influence the inertial, centripetal,
and Coriolis forces. Shi et al. [33] have developed a theoretical vehicle-bridge model to
investigate dynamic behaviour of slab bridges at various span lengths at different vehicle
speeds and road surfaces. Using finite deference method, Esmailzadeh and Ghorashi [34]
have studied the vibration of a Timoshenko beam under a partially dispersed moving load.
The response of the beam, bending moment and distribution of shear force are determined
in this analysis. Lee [35] presented the Lagrangian and the assumed mode approach for
analyzing the dynamic response of a Timoshenko beam. The study was carried out at
constant mass, speed and the beam slenderness ratio. Khalily [36] have studied the
dynamic behaviour of cantilever beam under moving mass. Mofid and Shadnam [37] have
developed an inexact method to describe the response of structures with internal hinges
and various end states under a traversing mass. Using the combined effect of finite
element and finite difference method, Cifuentes [38] has determined the response of a
beam vibrated by traversing mass. The methods used in the analysis are based on
Lagrange Multiplier. Nikkhoo et al. [39] have employed the most favourable control
Chapter 2 Literature Review
10
algorithm with a time changing gain matrix with displacement speed feedback to examine
the response of structure excited by moving mass.
Grant [40] has analysed the consequences of rotary inertial and shearing deformation on
the transverse vibration of a Timoshenko beam under a concentrated mass. Thambiratnam
and Zhuge [41] have developed a simple method for the dynamic analysis of structures on
elastic foundation due to the effects of moving load. The effects of moving mass speed,
the dynamic magnification on deflection, stresses and the foundation stiffness on the
response of the structure are determined. Rao [42] has determined the dynamic behaviour
of Euler-Bernoulli beam under a traversing load using mode superposition method. The
method of multiple scales was applied to solve the equation of motion of the time varying
mass systems. Using discrete element technique, the dynamic behaviour of Timoshenko
beam vibrated by travelling load was studied by Yavari et al. [43]. The response of a
double -beam structure has been studied under constant moving mass speed by Abu-Hilal
[44]. The effects of the traversing mass speed, the elasticity of the viscoelastic level and
damping of the beam are investigated on the dynamic responses of the beams. Wang and
Zhang [45] have investigated the vibration analysis of a guideway due to the effects of a
moving maglev vehicle.
Yang et al. [46] have presented a methodical study on the free and forced vibration of a
non-homogeneous beam subjected to axial force and traversing mass with the presence of
open edge cracks. Yan et al. [47] have studied the dynamic behaviour of functionally
graded beams on elastic foundation with open edge crack carrying moving mass.
Parametric studies with different boundary conditions of structures are conducted to know
the importance of crack depth, crack location, gradient of material properties, slenderness
ratio of the beam, moving speed and stiffness of foundation. Shafiei and Khaji [48] have
proposed a theoretical solution to study the free and forced vibration of a Timoshenko
beam with multiple open edge cracks under a moving concentrated load. They have
determined forced vibration response of the structure using the method of modal
expansion. Suzuki [49] investigated the consequence of acceleration on the dynamic
response of finite beam subjected to traversing load. Employing double Laplace
transformation, Hamada [50] determined the response of a damped Euler-Bernoulli simply
supported beam under the action of traversing concentrated force. Lee and Ng [51] have
studied the vibration response of a beam carrying a moving mass on one surface and
having a single crack on opposite side. Ichikawa et al. [52] have explored the dynamic
Chapter 2 Literature Review
11
behaviour of continuous structure carrying moving load using the concept of
eigenfunction expansion method. The influences of inertia and speed of moving load on
the response of the beam are described. Wu et al. [53] have employed both finite element
and analytical methods to study the response of a clamped-clamped beam under a
travelling mass primarily. The objective of this work was to predict the response of
moving crane structure using the above methods. Mallik et al. [54] have investigated the
steady-state response of an infinite beam on elastic support subjected to a traversing mass.
Aydin [55] have explained the characteristics of vibration of Euler-Bernoulli beams with
multiple cracks under the action of an axial load at numerous end conditions of the
structure. The effects of damages on buckling, edge cracks and axial loading on
eigenfrequencies and support conditions are discussed.
Sieniawska et al. [56] have detected the flexural stiffness of structure using the response of
traversing load. Yan and Yang [57] studied the flexural vibration of functionally graded
beams in the presence of open edge cracks under the action of both axial compressive
force and concentrated transverse load traversing along the beam. Using spectral element
method in time domain, Chen et al. [58] have presented the dynamic behaviour of
Timoshenko beam excited by an accelerating mass. Zarfam and Khaloo [59] have
investigated the response of structure on elastic foundation excited by traversing vehicle
and lateral vibration. Johansson et al. [60] have proposed a closed form solution to
examine the vibration analysis of a multi-span bridge structure carrying moving load. The
objective of the work was to carry out the vibration analysis of stepped beam under
constant traversing mass. Lou and Au [61] have developed finite element formulae to
evaluate the internal forces, shear forces and bending moment of Euler-Bernoulli beam
subjected to travelling mass. The discontinuities occurring due to the variation of shear
and internal forces in continuous beam are efficiently predicted using the developed
formulae. Museros et al. [62] have studied the free vibration analysis of a simply
supported structure under travelling mass.
Zarfam et al. [63] have explored the response spectrum of Euler-Bernoulli structure
subjected to time varying mass with the action of harmonic and earthquake support
vibration. Azimi et al. [64] have considered the effect of longitudinal acceleration of the
vehicle to formulate a numerical method for vehicle-bridge interaction dynamics. Cicirello
and Palmeriet [65] have studied the static vibration analysis of Euler- Bernoulli structure
with a random number of cracks under the action of both axial and transverse load. Zhong
Chapter 2 Literature Review
12
et al. [66] have analysed the dynamic behaviour of the prestressed bridge and moving
vehicle using vehicle-bridge interaction dynamics. Costa et al. [67] developed a theoretical
formulation for the critical speed of traversing mass problem on elastic foundation. Fu
[68-69] has given the attention to the instant of switching of cracks and its effects on the
dynamic excitation of continuous bridge structure under travelling vehicles. In the later of
his investigation, he has formulated a numerical solution for a cracked simply supported
bridge with switching cracks under the excitation of seismic and the moving vehicle. Aied
et al. [70] have applied the ensemble empirical mode decomposition (EEMD) to study the
response of acceleration of a bridge structure subjected to moving load with the intention
of confining immediate change of bridge stiffness.
2.2.2 Finite element analysis/method for response analysis of structure
Using finite element method, Hino et al. [71] have determined the deflection and
acceleration of a reinforced concrete bridge structure under traversing vehicle load. The
analysis has been carried out with constant magnitude and speed considering the bridge
over River Bramphaputra in India. Bhashyam and Prathap [72] have presented the
efficiency of Galerkin finite element method to examine the large vibration amplitude of
fixed structures with different end conditions. Olsson [73] has formulated a method to find
the response of bridge with the action of moving load in modal coordinates. He has
established some of the finite element methods for traversing load problems. Yoshimura et
al. [74] have investigated the random excitation of the non-linear structure with different
types of sectional areas under a moving vehicle. The longitudinal and transverse
deflections of the beam are calculated using the Galerkin finite element method. Lin and
Trethewey [75] have presented a finite element formulation for structures under different
types of moving mass. Chang and Liu [76] carried out the random vibration analysis of a
nonlinear structure on an elastic foundation under traversing load. The deterministic and
statistical responses of the structure have been calculated by combining Galerkin method
and finite element analysis.
Rieker and Trethewey [77] have explored the finite element analysis of elastic structures
subjected to traversing distributed load. This work was modelled for the improvement of
the distributed railway train mass structure. Song et al. [78] have developed a novel finite
element model for three-dimensional finite element analyses to examine the response of a
structure subjected to high-speed trains. The objective of their work was to develop better
finite element models to be used in structural element railway bridge structure. The
Chapter 2 Literature Review
13
developed finite element model was verified through numerical examples considering a
simply supported steel-concrete and a PC box-girder bridge structure. Law and Zhu [79]
have examined the response of cracked reinforced concrete structures under moving
vehicles in the presence of both open and breathing cracks. Experimental procedure has
been carried out on a Tee-section beam subjected to moving vehicles to ensure the
significance of crack models. The dynamic bridge deflection, change in relative and
absolute frequency, and phase plot of the dynamic responses of structures are studied for
the probable correlation of crack modelling as open or breathing crack.
Using the modal superposition and closed form solution method, Yang and Lin [80] have
determined the transverse deflection for a bridge and traversing vehicle through vehicle
bridge interaction dynamics. The deflection, speed and acceleration of the moving
structural system are calculated by the fundamental frequency of the bridge and the
driving frequency of the moving vehicle. Using the commercial software LS-DYNA, a
review of finite element analysis procedure was carried out by Kwasniewski et al. [81] to
study the vehicle-bridge interaction systems. The experimental test was conducted in a
bridge structure at Florida, US.
Ju and Lin [82] have developed a finite element model to explore the responses of vehicle-
bridge dynamics due to the application vehicle braking and acceleration. To verify the
finite element method for the effects of vehicle braking and acceleration, a two axle
vehicle traversing on bridge structure was formulated through semi-theoretical solution.
The application of this proposed method is limited to linear and small displacement
analysis. Bajer and Dyniewicz [83] have proposed an effective space-time finite element
method to solve the problems based on moving load. Kahya [84] has presented a multi-
layer shear deformable composite beam structure subjected to traversing mass using finite
element method. The consequences of moving speed and laminated lay-up on the response
of the beam structure with different end conditions have been determined. He has
concluded that angle-ply laminates and laminas stacking have a significant influence on
the beam response. Dyniewicz [85] has proposed the application of velocity formulation to
explain general moving inertial mass problem using space-time finite element method.
Esen [86] has developed a novel finite element method to examine the vertical vibration
of a rectangular plates structures excited by traversing load. Zhang et al. [87] have
formulated a finite element model for bridge structure using the equivalent orthotropic
material modelling (EOMM) for the application of multi-scale dynamic loading. The
Chapter 2 Literature Review
14
dynamic responses and properties of a simplified bridge structure are attained using the
equivalent orthotropic material modelling method. Amiri et al. [88] have analyzed the
vibration based dynamic behaviour of a Mindlin elastic plate under the excitation of
traversing mass using the method of eigenfunction expansion. The first order shear
deformation hypothesis was applied to determine the response of plates subjected to
moving load excitation with different end conditions. Ju [89] investigated the
improvement of bridge structure for the safety of travelling trains during earthquakes
using finite element analysis. The interaction and separation of rails and wheels were
accounted in this analysis. Nejad et al. [90] have predicted the natural frequencies and
modal shapes of a double cracked beam in the theoretical formulation using Rayleigh's
method. This method has better applicability over eigenvalues method. But the accuracy
of this method is limited to small crack depth. Aied and Gonzalez [91] have explored the
response of a simply supported structure subjected to traversing load and determined the
deviation of strain rate and its consequence on the modulus of elasticity. The significance
of mass magnitude and speed on the dynamic deflection and strain rate of the structure are
also examined.
Wu et al. [92] have studied the dynamic behaviour concrete pavement structure subjected
to traversing load. The concrete pavement structure has an asphalt isolating layer on its
surface. The response of the structure has been carried out by finite element method using
the commercial ABACUS software. Stress and dynamic deflection of the concrete
structure are also determined at the critical moving load position by altering the depth,
modulus of elasticity of isolating layer and the amalgamation between the dividing layer
and concrete slab. Jorge et al. [93-94] have analysed the dynamic behaviour of structure
supported on nonlinear elastic foundation under travelling mass. The consequences of the
intensity of travelling load and speed and foundation’s stiffness on the response of the
structure have been investigated. The objectivity of this work is to study the structural
behaviour on any types of realistic foundation. Further, they have extended their research
work on the analysis of moving load structure problem on a bilinear foundation. The
critical speed of the moving load and deflection of the structure UIC-60 rail systems are
determined in finite element domain. Alebrahim et al. [95] have studied the dynamic
excitation of a self-hauling composite structure subjected to traversing load using transfer
matrix method and finite element method.
Chapter 2 Literature Review
15
Ozturk et al. [96] have carried out the dynamic analysis of a hinged-hinged damaged
structure cracked on elastic foundation subjected to traversing load using finite element
method. The dynamic response of the structure has been calculated using Newmark-
integration method in finite element domain. The consequences of crack depth and
position, traversing mass speed, the elastic foundation on the beam dynamic deflection
have been verified. Fu [97] has examined the dynamic vibration of a simply supported
bridge structure with the existence of switching cracks under the action of traversing train
load and seismic vibration. The vibration of the bridge structure was studied by modal
analysis method in finite element domain. From the analysis of his work, he has remarked
that the switching crack can alter the stiffness of the structure due to seismic vibration, and
the deflection of the structure can amplify. Using finite element method, Beskou et al. [98]
have investigated the effects of 3-D pavement under traversing vehicle excitation. The
FEM was carried out in the time domain using the commercial ANSYS program for the
response of the vehicle-structure interaction analysis. They have observed that with the
increase in moving speed, the dynamic response of the pavement structure also increases.
Their work has been limited in linear elastic material properties.
2.3 Damage detection
2.3.1 Classical/FEA based methods for damage detection in structure
Dutta and Talukdar [99] have proposed a damage detection technique using the alteration
in dynamic response between the undamaged and damaged states. Eigen values analysis
has been carried out by employing the algorithm of Lanczos in adaptive h-version in finite
element domain to control the discretization fault to evaluate the modal parameters
precisely. Friswell and Penny [100] have compared various methods of crack modelling to
structural health monitoring problem using the low-frequency vibration and crack
flexibility models based on beam elements. The effects of breathing cracks on vibration
are also investigated on the bilinear stiffness of beam elements. Lee et al. [101] proposed a
damage identification method for bridge structures subjected to a passage of vehicle
loading using the data of ambient vibration. Abdo and Hori [102] have carried out
numerical formulation fault detection in a structure using the relationship between cracks
and dynamic properties changes. The damaged region has been found out using
characteristics of rotation of mode shape in this analysis. All these studies are carried out
in finite element analysis domain. A novel algorithm for crack detection and quantification
Chapter 2 Literature Review
16
in the structure are formulated by Kim and Stubbs [103] using the alteration in modal
characteristics. Their algorithm was applied to detect the crack in two-span continuous
structure with the knowledge of pre and post-crack modal parameters. Chondros et al.
[104] have explored a continuously damaged beam vibration theory for the transverse
vibration of damaged beam with the presence of single and double edge open cracks.
Chondros and Dimarogonas [105] have studied the vibration analysis of cracked cantilever
beam with single edge open crack. The analysis has been further extended for the
detection of multiple cracks in the structure. Khoshnoudian and Eafandiari [106] have
developed a damage diagnosis method using the modal data and finite element analysis of
the structure. Swamidas et al. [107] have predicted the significance of crack size and
location of the fundamental frequencies of the structure using the Timoshenko and Euler
beam formulations theory. The consequences of cracks on shear deformation and rotary
inertia of the structure are also examined. Mazurek and DeWolf [108] conducted
experimental analysis to predict the possible damages in bridges using the vibration
signatures. Important efforts were given on automation, acquisition and study of vibration
signatures for up gradation of a bridge structure. Ruotolo and Surace [109] explored a
fault detection technique for a structure with multiple cracks using the modal data at the
inferior modes. Using the changes in flexibility, Pandey and Biswas [110] proposed a
damage identification method for structures and verified the method through experimental
tests. Using the Frobenius method, Chaudhari and Maiti [111] have examined the
vibrations analysis of a geometrically segmented slender structure in the transverse
direction with and without a crack. Considering the natural frequency as an input, they
have applied this method for crack detection and sizing. The problem identifying stiffness
changes in bridge structure under traversing oscillator have been proposed by Majumder
and Manohar [112] in time domain analysis. Chinchalkar [113] has developed a
computational method to determine the crack location in beam structure employing the
first three natural frequencies of the beam by finite element analysis.
Haritos and Owen [114] employed the vibration data for the assessment of damages in
bridge structures. The damaged identification is carried out using system identification and
statistical pattern recognition methods individually in a reinforced concrete bridges. They
have concluded that these two methods are to be complementary to each other and a good
approach for structural health monitoring problems related to bridges. A theoretical along
with experimental method has been employed for crack detection in the beam by Nahvi
Chapter 2 Literature Review
17
and Jabbari [115]. This method was based on the measurement of natural frequency and
mode shapes. Alvandi and Cremona [116] have reviewed various structural fault detection
methods like flexibility change method, flexibility change curvature method, mode shape
curvature method, and strain energy method based on vibration signatures. From the above
fault detection methods, strain energy method is the effective one. Wang et al. [117] have
presented damaged detection method in structure named Local Damage Factor (LDF)
which can predict the existence, location and quantification of damage. For the
verification of the proposed method, a practical study was carried out in a 3-D steel frame
and wharf. Fushun et al. [118] have proposed a fault identification method in a bridge
subjected to a moving vehicle by introducing a method named ‘moving load damage-
locating indicator (MLDI)’. The values of modal curvature and MLDI at every node of the
baseline of the structure and the damage models have been calculated. The sudden
possible changes of MLDI values would give the possible location of the damage. Yu et
al. [119] have carried out a review analysis in the present information on factors
influencing the performance and identification of traversing loads in bridges.
Sekhar [120] has conducted various studies to examine the influences of cracks on a
structure and summarized different crack detection methods using vibration signatures.
Based on bending vibration measurement, a theoretical method for crack detection in
uniform structures with different boundary conditions with the presence of single edge
crack has been proposed by Khaji et al. [121]. The developed theoretical method has been
verified through numerical examples. Using the vibration amplitudes, Lee [122] developed
a method to identify multiple cracks in the beam by finite element method. The developed
method has been validated by Newton-Raphson method numerically. Zhu et al. [123] have
presented a novel technique to identify cracks in large-scale concrete columns for the
assessment of automated bridge conditions. The results from the developed technique have
been compared with manual detection of faults to check the precession of the method.
Zhang et al. [124] have presented virtual distortion method for instantaneous detection of
traversing mass and damages in structure from the dynamic response of the structure.
They have used the forces between the traversing mass and structure as excitation forces.
An analytical as well as experimental verification, have been conducted to identify the
crack in a simply supported beam using the natural frequencies of the cracked beam by
Sayyad and Kumar [125]. A relationship among location and size of crack and natural
Chapter 2 Literature Review
18
frequencies of the beam has also been formulated. The analysis has been carried out in
finite element analysis and verified by experimentation.
Dilena and Morassi [126] have presented dynamic tests on a damaged bridge. The
deviations of modal properties at lower modes of vibration after enforcing the artificial
incremental damages are determined using the harmonic force tests. Roveri and Carcaterra
[127] have proposed an innovative method based on Hilbert–Huang transformation to
detect damages in bridge structure under traversing vehicle load. The forced response of
the structure has been found out by modal analysis method and the single-point response
obtained by the Hilbert–Huang transformation. The possible location of damage is
exposed by the assessment of the first instantaneous frequency curve which creates a
quick crest in the association of the damaged segment. Nguyen [128] has carried out a
comparison analysis for crack detection in structures under moving load with on open and
breathing cracks. The beam stiffness with an open crack has been determined from the
concept of fracture mechanics where as beam stiffness with a breathing crack has been
modelled as a time-dependent matrix based on the stiffness of the beam with open crack.
Li and Law [129] have formulated a substructural damage detection method subjected to
travelling vehicles using the dynamic response reconstruction procedure. The finite
element modelling of the undamaged structure and the determined dynamic acceleration
response from the damaged substructure are necessary for the damage identification in the
structure. Zhu and Law [130] have formulated a damage identification procedure of a
cracked simply supported structure under travelling load in the time domain using the
interaction forces between the travelling load and structure as the vibrating force. The
interaction forces between the vehicle-bridge and damage of the structure on the bridge
deck are recognized from the calculated responses of the damaged bridge through
succession iterations without the information of the moving loads.
Khiem and Lien [131] investigated multi-cracked identification problem for one-
dimensional structure using natural frequencies. Patil and Maiti [132] have developed a
crack detection method to predict the position and quantification of cracks in a slender
cantilever beam. The method has been formulated using the natural frequencies of the
beam and later verified through experimental tests. Pakrashi et al. [133] developed an
experimental investigation of cracks in bridges using the vehicle-bridge interaction forces.
This analysis could be helpful for the online health monitoring problem of bridges under
moving vehicle. Zhou et al. [134] have developed a magnetic wire smart film for crack
Chapter 2 Literature Review
19
detection in large-scale concrete bridges. The smart film inventions, the basic operational
principle of the film, and the film application on an actual bridge structure have been
explained. Bouboulas and Anifantis [135] have developed a finite element model to
analyze the behaviour of a vibrating beam with the presence of non-propagating edge
crack. The response of the beam is studied using either Fourier or wavelet transformation
to evaluate the effects of breathing cracks. The consequences of the angle of cracks, crack
depth and position, are investigated on the vibrational behaviour of the cracked structure
through various parametric studies. Zhan et al. [136] have proposed a fault isolation
method for railway bridges by utilizing the train-induced dynamic responses and
sensitivity investigation. The significance of measurement of noise and track irregularities
is explained in the proposed method.
Ghadami et al. [137] have presented an easy method to identify, locate and quantify
multiple cracks in the structure using the measured natural frequencies. The key
objectivity of the proposed method is to identify the unknowns of cracks intervention. Hu
and Liang [138] have proposed an integration method based on the theory of vibration to
detect arbitrary number of cracks in structure. The utilization of massless insignificant
springs and the continuum of damage conceptions have been integrated to locate the
potential damages. Chomette et al. [139] have investigated an innovative method to
identify tiny cracks on structure using active modal damping and piezoelectric mechanism.
The difference in active damping is identified using the Rational Fraction Polynomial
method as a pointer of cracks detection in the proposed method. Nejad et al. [140] have
approximated a theoretical method by extending Rayleigh's method for a structure with
single or double cracks to determine the natural frequencies and mode shapes of that
structure. Wang et al. [141] have developed a fault diagnosis system in beam type
structure using the statistical moment analysis. Numerical analysis has been carried out on
a damaged simple supported and two-span continuous structure to show the accuracy of
the formulated method. Using the finite element method, Nguyen [142] has analyzed the
mode shapes of a damaged beam and applied it for damage identification. The significance
of the coupling system between longitudinal and transverse bending vibrations due to the
presence of crack on the mode shapes has been inspected. This quantitative mechanism
has also been employed to estimate the size and location of the crack with the pragmatic
coupled modes.
Chapter 2 Literature Review
20
Lee and Wu [143] proposed a simplified method to replicate multi-cracks in structure.
Local adaptive differential quadrature method has been developed to find out the natural
frequencies of the beam. The position and depth of cracks have been predicted using the
Newton–Raphson iteration analysis. Two crack transfer matrix has been investigated in
finite element method for prediction of cracks in a structure by Nandakumar and Shankar
[144]. The developed damage identification procedure was also verified through
experimental tests using the local detection of sub-structure of a fixed beam. This method
is also applicable for damage detection in large structures. Khiem and Tran [145] have
investigated a crack isolation method for structure with arbitrary number of cracks using
vibration mode. The crack detection and localization have been done effectively with this
method from the sparsely and noisy data also. Ettefagh et al. [146] have designed an
innovative damage identification algorithm based on method of model updating. The
design method has been employed to replicate the deteriorated structure considering
various levels of noise to check the efficiency of this method. Schallhorn and Rahmatalla
[147] have proposed a vibration based damaged identification procedure to access the
health monitoring condition of a highway steel-girder bridge.
He and Zhu [148] have explored a closed form solution to analyse the response of a
deteriorated simply supported structure under traversing mass and investigated the
consequences of local stiffness loss based on the traversing frequency element
corresponding to the traversing load and the natural-frequency element of the beam. A
deep insights into structural damage identification based on the traversing mass-induced
dynamic response has been developed using this method. The consequences of altering
traversing speed and traversing load dynamics on damage localization are also explained.
Zhu et al. [149] have determined the local damage of the large and intricate structure
subjected to traversing load. Law and Lu [150] have proposed a time domain analysis to
identify crack parameters from the measurement of beam deflection. The response of the
moving load structure is calculated by modal superposition method. The presented damage
detection method has been proved with experimental works through impact hammer test
on a structure with single crack. Using alteration in the nonlinear characteristics structures,
Law and Zhu [151] have developed a damage detection procedure of a reinforced concrete
beam structure subjected to moving load. Nonlinearities characteristics of the beam
structure have been identified by considering the alteration in the instantaneous frequency
at different positions of the vehicle load along the beam. Bu et al. [152] have proposed a
Chapter 2 Literature Review
21
novel approach to assess the damage occurring in a bridge deck using the measured
dynamic response of a moving vehicle on a structure. The vehicular load has been treated
as a smart sensor and force transducer in the moving load interaction structure to detect
the damages.
Karthikeyan et al. [153] have developed a method to know the location and quantification
of cracks in a structure using the modal parameters in finite element model. A vibration
based damaged detection method has been recommended by Zhou et al. [154] to estimate
the position and quantification of damages on a bridge deck. The analysis has been carried
out in finite element analysis as well as laboratory test experimentation. Based on
mathematical modelling, Al-said [155] has intended a crack detection method to locate
and quantify cracks in a stepped beam under a slowly traversing mass. In the proposed
method, the crack is identified by monitoring a single natural frequency of the system. The
validation of the method has been done with experimental test along with finite element
analysis. Yin and Tang [156] have carried out a computational study to detect multiple
cracks in a cable-stayed bridge using the transverse deflection of a vehicle crossing over a
bridge. Li et al. [157] have performed a numerical analysis along with experimental
verification to estimate damages in a structure under a moving load. Experimental analysis
has been conducted on a Tee-section concrete structure subject to a moving vehicle for the
accuracy of the proposed method. Li and Hao [158] have used the relative displacement
measured from the vibrant response of the passing vehicle on the structure as damage
indicator. Using the vehicle transmissibility of vehicle-bridge coupled structure; Kong et
al. [159] have studied a fault diagnosis method for bridge structure under a passing
vehicle. Various types of damage indicators are prepared artificially and based on the
transmissibility of the response of the vehicle and bridge structure, and a damage detection
procedure has been conducted numerically.
Aydin [160] has recommended an effective procedure to determine the vibration
characteristics of a flexible structure subjected to axial load. Dar et al. [161] have
performed an experimental study on the dynamic behaviour of a deteriorated bridge
structure. The objectivity of the work was to find the load-carrying capacity of the bridge.
Oshima et al. [162] have proposed a damage assessment method for bridge structure using
the estimated mode shapes due to the traversing vehicle. The proposed method has been
proved by various numerical solutions. Mazanoglu [163] has recommended crack
detection procedure for multiple cracks based on the concept of measurement of natural
Chapter 2 Literature Review
22
frequency ratios. The procedure was verified through frequency ratios obtained from the
experimental test. Feng and Feng [164] have proposed a damage isolation procedure for
bridge structure by utilizing the vehicle induced deflection without the knowledge of the
traffic-induced vibration and surface roughness. The probability and performance of the
developed method have been validated by numerical simulation with single and multiple
cracks.
2.3.2 AI techniques based methods for fault detection in structure
2.3.2.1 Genetic Algorithm based methods for crack detection in structure
Chou and Ghaboussi [165] have formulated an optimization problem using the genetic
algorithm (GA) for damage detection in structure. The deflections at unmeasured degrees
of freedom have been calculated by GA to avoid the analysis of structure in fitness
estimation. Shopova et al. [166] have solved various types of optimization problem using
GA. A common genetic algorithm problem has introduced the basic principle of the
number of selection, reproduction and mutation parameters in the correspondence, genetic
operators. He et al. [167] have recommended a bridge damage detection technique using
the vehicle induced bridge response as a direct problem and genetic algorithm as a reverse
problem. The genetic algorithm is employed to identify the pattern and location of
damage. Sun et al. [168] have designed a vehicle suspension system using the genetic
algorithm and chosen the minimum dynamic pavement moving load as the design
principle. Baghmisheh et al. [169] have proposed a fault isolation method for cracked
structure based on genetic algorithm. A Genetic algorithm has been used to examine the
natural frequencies of the cracked structure. The position and depth of cracks in the
structure are identified by the continuous genetic algorithm optimization method. Meruane
and Heylen [170] have developed a hybrid coding based on genetic algorithm to identify
the cracks parameters in structures. The proposed method can also identify the happening
of false damage due to the errors in the noisy experiment and numerical simulations.
Buezas et al. [171] have employed a finite element method along with a formulated
genetic algorithm for fault diagnosis in damaged structure. The formulated optimization
method can identify not only beam like structure but also structure like curved beam and
blade like structural element with rotational motion.
Na et al. [172] have investigated a damage detection procedure for a shear building using
genetic algorithm followed by the flexibility matrix with the dynamic response. The
Chapter 2 Literature Review
23
proposed method has been exemplified with numerical analysis. Mehrjoo et al. [173] have
applied the genetic algorithm as inverse method to estimate the location and severity of
cracks in cracked Euler-Bernoulli beam. The inverse method has been verified with
different types of damage scenarios. Lee [174] has applied the finite element method as
direct method and coupled genetic algorithm as the reverse method to detect the traversing
mass on the bridge deck. This method can also identify the mass of the moving load as
well as the traversing speed. Chisari et al. [175] have developed a fault isolation method
for a base-isolated and post-tensioned concrete bridge using the genetic algorithm. The
algorithm was based on static and dynamic loading condition.
2.3.2.2 Neural network based methods for crack detection in structure
Alli et al. [176] developed a method based on artificial neural network (ANN) to solve the
dynamic systems problems. The developed method has been applied for the solution of
vibration control problem. Cao et al. [177] have recommended a method to detect the
flight loads on aircraft wings based on ANN approach using the relationship of load-strain
analysis in structures. Mahmoud and Kiefa [178] have explained vibration problem on
cracked structure employing general regression neural networks (GRNN). The accuracy of
the employed method has been checked with a cracked cantilever beam with an edge
crack. The location and size of the crack have also been identified. Waszczyszyn and
Ziemianski [179] have presented some fundamental concepts on back-propagation neural
network (BPNN) to study the bending effects on elastoplastic structure, problems on plain
stress, the fundamental concept on vibration analysis of real building and damage
identification. Zang and Imregun [180] have jointly approached the ANN and principal
component analysis (PCA) for structural damage detection by reducing the frequency
response functions (FRFs). The FRFs have been used as the input parameters to the
network and damage parameters as output.
Liu et al. [181] have employed a BPNN and computational mechanics to identify cracks in
the structure. The types of cracks present, the extent of cracks, location and severity of
cracks are evaluated using the approached BPPN. Chen et al. [182] have developed a fault
isolation method of structure based on neural network using the response data and
transmissibility function as input to train the proposed network. Kao and Hung [183] have
investigated a neural network based damage detection procedure using the changes in
properties of unknown structural systems. This procedure involves two systems namely
system identification which detects the uncracked and cracked states of structural systems
Chapter 2 Literature Review
24
and structural damage detection. Sahin and Shenoi [184] have presented a method to
quantify and localise the damages in a beam like structure using ANN and verified it
through experimental validation. The alteration in first three natural frequencies and
curvature mode shapes of the structure obtained from FEM have treated as the input
parameters for the proposed ANN and in the later part, the location and quantification of
the cracks have been predicted by the proposed ANN.
Ataei et al. [185] have applied the linear two layer feed forward neural network (FFNN)
with back propagation learning technique to measure the stain and deflection from railway
bridge load test. Kang et al. [186] have proposed a BPNN to estimate the fatigue life of a
structure subjected to multi-axial loading. The prediction of the fatigue life period was
based on critical plane concept using finite element modelling. Yan et al. [187] have
summarized various methods for structural damage detection based on vibration
signatures. It covers the theory of intelligent damage isolation method and its application
in prediction in structural damage recognition. Li and Yang [188] explored an innovative
technique for structural damage diagnosis using ANN with arithmetical properties of
structural dynamic induced responses. The alterations of variances or co variances of
responses of structure are chosen as damage parameters for damage isolation. The BPNN
with the alteration of variance of responses of structure as input parameters and damage
index as output parameters are applied for damage isolation in structure. Mehrjoo et al.
[189] investigated a method for prediction of the crack intensities of joints of truss bridge
structure by applying BPNN. The exactness and effectiveness of the adopted BPNN
method was exemplified with numerical analysis. Bakhary et al. [190] approached a
method to estimate petite structural damage based on ANN in association with multi-stage
sub-structuring. The position and extent of the damages are identified by substructure
method with multi-stage ANN modelling. The approached method was experimentally
verified by considering continuous concrete slab and a three-storey portal frame structure.
Al-Rahmani et al. [191] have investigated a combined approach for fault detection in
bridge structure using soft computing mechanism. The mechanism was based on
ABAQUS finite element analysis and ANN methods. Using the arithmetical properties
obtained from the dynamic response of a moving train, Shu et al. [192] have developed
ANN-based method to diagnose structural defects in a simplified railway bridge structure.
The analysis has been carried out for a single span bridge. Yaghinin et al. [193] have
proposed a hybrid algorithm based on the application of the training of the ANN and
Chapter 2 Literature Review
25
explained the exactness, different mechanism and validation of the developed algorithm.
Elshafey et al. [193] have estimated the crack spacing in a concrete structure using neural
network. Erkaya [194] has predicted the vibration characteristics of a gear system using
neural networks. The proposed network has been trained with Levenberg–Marquardt
learning mechanism. Pandey and Barai [195] have developed the method of multilayer
perceptron learning to identify multi-damages in bridges. The developed method was
applied to a truss bridge structure. Karninsk [196] proposed a method using ANN as a
predictor to localize the damage employing the change of natural frequencies as inputs.
This method has been applied to a free-free beam for the exactness of the ANN model.
Seibi and Al-Alawi [197] estimated the toughness of fracture using ANN and analyzed the
consequences of crack geometry, temperature and toughness of fractures. Zhao et al. [198]
have proposed a computational intelligence based NN method to diagnose faults in
various types of structures like a beam, frame and support movement of beams. The
requisite data to train the ANN model were obtained from FEA. Marwala and Hunt [200]
have employed the frequency responses and modal parameters obtained from FEA and
ANN to estimate faults in damage vibrating structure. Chang et al. [201] have proposed an
iterative ANN model for structural damage identification. An adapted back- propagation
mechanism has been developed to estimate the damage parameters and verified with the
experimental investigation. Lin et al. [202] have investigated an NN- based method to
estimate damages of bridges during major earthquakes. The proposed method was applied
to bridges in Taiwan to predict the seismic damages. Zang et al. [203] approached
combined efforts of independent component analysis and ANN for structural fault
diagnosis. The exactness of the proposed method was verified in real life study
considering truss and bookshelf structures. Yeung and Smith [204] investigated a crack
detection procedure using neural network by employing vibration signatures obtained
from FEA. Using the errors occurred in baseline FE model, Lee et al. [205] explored fault
diagnosis procedure for the bridge under passing vehicle. The developed method has been
verified in a laboratory test considering a simply supported and multi-girder bridge
structure. Bakhary et al. [206] have proposed an NN model based on statistical properties
to detect damages in structure considering the uncertainties occurred in the NN model.
The proposed approach was verified experimentally with a steel portal frame and concrete
slab model.
Chapter 2 Literature Review
26
Wong et al. [207] developed an algorithm based ANN for online identification of damages
to structure induced due to ground shaking. A 2-story steel-frame building was used to
check the accuracy of the approached algorithm. Pawar et al. [208] applied spatial Fourier
analysis along with NN to analyse faults in structures. The approaching methods were
verified through numerical examples with the presence of noises. Gonzalez-Perez and
Valdes-Gonzalez [209] presented ANN based structural damage detection technique to
bending in the girders of bridge subjected to vehicle loading. The differences in modal
strain energies are considered as inputs to the ANN model with 12,800 damages scenarios
in the proposed technique. Li [210] proposed a probabilistic neural network approach for
localization of defects in the composite plane structure. Parhi and Dash [211] have carried
out finite element modelling along with neural network based damage prediction to
identify multiple cracks in the beam-like structures using the vibration signatures as inputs
to the proposed ANN model. The feed-forward multi-layered neural network based on
back-propagation mechanism was applied for the estimation of crack parameters. Elshafey
et al. [212] have estimated the crack width in a concrete structure by ANN. The radial
basis and feed-forward neural networks with back propagation algorithm were applied to
predict the crack width in the structure. Hasancebi and Dumlupınar [213] have applied the
nonlinear finite element modelling along with ANN to analyze the load rating of bridge
structures. The 3D FE model along with experimental verifications has been carried out to
determine the nonlinear response and load rating of single span T-beam Bridge.
Using frequency response function, Bandara et al. [214] have applied ANN method to
identify defects in a structure for a given label of vibration. The key objective of the
proposed work is to analyse a feasible technique for structural health monitoring based on
vibration responses which would reduce the elements of the initial frequency response
functions. Aydin and Kisi [215] have applied the Multi-layer perceptron and radial basis
neural networks based mechanism for fault isolation in the beam-like structures. The
modal properties of the structures as input parameters have been provided to the proposed
networks. It has been shown that the radial basis neural networks have performed better
and can be used as damage identification algorithm. Kourehli [216] has approached an
innovative idea for fault detection in structures using the incomplete modal data and ANN.
The ANN model has been trained by the first two incomplete mode shapes, and natural
frequencies of the structure obtained from the finite element model with feed forward
back-propagation mechanism. The accuracy of the model was validated with a three-story
Chapter 2 Literature Review
27
plane, spring-mass and simply supported structures. Alavi et al. [217] have integrated the
finite element method along with probabilistic neural networks with the concept of
Bayesian decision mechanism to identify damages in structures. The exactness of the
innovative technique was primarily estimated in a simply supported structure under three-
point bending and later in a bridge gusset plate.
2.3.2.3 Recurrent neural network based methods for crack detection in structure
Yu et al. [218] have applied Elman’s recurrent neural network (RNN) to estimate the
performance of a boring mechanism through its full life cycle. Xiong and Withers [219]
have proposed an RNN model to predict the progression of the damages produced during
hot non-uniform and non-isothermal forging processes. The hyper parameters related to
the noise level and weight decay of the proposed RNN model has been trained with
introducing the Bayesian algorithm. Ekici et al. [220] developed an Elman’s RNN based
method to estimate the locations of transmission line fault. Wavelet transformation
method was implemented to select characteristic description about the faulty signals. The
developed ERNN model can predict the fault locations quickly and an alternative
characteristics to the feed forward back propagation networks and radial basis functions
neural networks. Abdelhameed and Darabi [221] have applied the RNN based mechanism
to control the fault-tolerant of mechanized in order manufacturing systems under sensor
faults. The training procedure of the applied RNN model has been performed based on
training data produced from the well-mechanized system run by a programmable logic
controller. Connor et al. [222] have developed a healthy training mechanism and applied it
to RNNs. The proposed mechanism was based on filtering outliers from the induced data
and predicted the required parameters from the filtered data. Pearlmutter [223] has
determined the gradient of a dynamic RNN using back propagation algorithm. Tse and
Atherton [224] have predicted the deteriorated conditions of machine structures using
vibration parameters and RNN. The significance of defects and remaining life span of
machine structures are also estimated applying the proposed method.
Gan and Danai [225] have proposed a rule based RNN to model dynamic systems. The
rule based RNN model has been designed on linearized state space modelling of the
dynamical system. Waszczyszyn and Ziemianski [226] have applied various types of
mechanics to ANN and RNN model to analyze faults in different types of machine
structures using back propagation algorithm. Valoor et al. [227] have developed a self-
adaptive vibration control structure using RNN to control the excitation of beam
Chapter 2 Literature Review
28
structures. Seker et al. [228] have used ERNN for the health monitoring of nuclear power
plant and rotating machines. Schafer and Zimmermann [229] have proved the RNN as
common approximators and showed its capacity in state-space model. Thammano and
Ruxpakawong [230] have introduced the concept of additional weight to the standard
JRNNs and ERNNs. Coban [231] has proposed a rule based context layer locally RNN to
identify faults in dynamic systems using back propagation mechanism. The accuracy of
the proposed model has been verified with the experimental application with a D.C motor.
2.4 Statistics based method for fault identification in
structure
Sohn et al. [232] have employed the statistical process control approach to structural
health monitoring problem. The approach was exemplified by replicating online
monitoring of damages in structures. Fugate et al. [233] have proposed a statistically based
damage isolation method using vibration signatures. The measured data are obtained from
the acceleration-time histories of undamaged structure based on the autoregressive model.
Residual errors which differentiate between the predictions from the autoregressive model
and the real data measured from the time- history are implemented as damage responsive
characteristics to the structures. Lei et al. [234] have developed a statistics based fault
diagnosis analysis using the time series prediction method on vibration signatures. Lu and
Gao [235] proposed an innovative time-series model to diagnose damages in structures.
The proposed model was originated in an exogenous form containing the acceleration
responses.
Mattson and Pandit [236] developed a mechanism based on vector autoregressive (ARV)
model to identify damage locations in vibrating structures. The existence and location of
damages are indicated by the standard deviation obtained from the residual data of the
ARV. Zhang et al. [237] proposed an inventive damage isolation mechanism using the
statistical moments of dynamic analysis of a structure subjected to arbitrary excitations.
The sensitivity of the proposed method to structural damage has been examined for
different types of structural responses and various orders of the statistical moment. The
method has also been extended from SDOF to MDOF systems. Gul and Catbas [238] have
carried out theoretical analysis followed by experimental verification for structural health
monitoring problems using time series modelling based on Statistical pattern recognition
method. The data obtained from ambient vibration are used to explain the proposed
Chapter 2 Literature Review
29
methodology. Law and Li [239] have assessed the condition monitoring of concrete bridge
structure by estimating the reliabilities of the structure. The mean and the standard
deviations values of the estimation results of the structure are determined and then
incorporated in the reliability analysis of the structure. The approach has been verified
with FEA considering a concrete bridge under the action of a moving vehicle.
Zapico-Valle et al. [240] have developed an inventive damage isolation procedure for a
cracked cantilever beam based on statistical process control-based method and verified
with experimental analysis. The concept of signal length has been established as the
characteristic for statistical process control. Kwon and Frangopol [241] have predicted the
life period performance estimation and management of aging steel bridge structures under
damages by combining reliability model, crack growth model and probability of detection
model as prediction models. Mosavi et al. [242] have developed an autoregressive model
to identify the crack location in structures using the vibration response data. Cavadas et al.
[243] have applied the data-driven method on traversing-mass responses to identify
damage existence and location in structure. The proposed data-driven method consists of
two components namely moving principal component and robust regression analysis for
the condition monitoring of the structure. A novel statistical time series mechanism with
the functional model have been developed by Kopsaftopoulos and Fassois [244] to
identify, locate and estimate damages based on vibration induced response of the structure.
Phares et al. [245] have conducted field verification on bridge damage isolation relied on
statistically based algorithm. The method can precisely identify the damage of the
structure using the strain data accumulated from various sensors on the bridge. Wang et al.
[246] have developed a two step method using the concept of statistical moment for crack
identification in beam type structure. Yu and Zhu [247] have applied the time series
analysis and the higher statistical moments based structural responses to identify nonlinear
damages in structures. Reiff et al. [248] have developed a statistical bridge damage
detection procedure employing the girder allocation factors under moving load excitations.
This work has been designed to alert the conditions of bridges and present a better
probabilistic approach to assess the damage conditions and localizations.
2.5 Miscellaneous methods
Masciotta et al. [249] have approached a fault isolation method using the second order
spectral characteristics of the nodal response of structures. The precise dependences on the
Chapter 2 Literature Review
30
frequency content of the outputs of power spectral densities are used as suitable
parameters for the identification and localization of damages. The method has been
validated through computational simulation considering the Z24 Bridge in Switzerland.
Dilena et al. [250] applied the interpolation damage isolation method for health
monitoring of structures based on frequency response function. The analysis has been
carried out for a single span bridge with growing levels of intense damages. Gokdag [251]
has investigated a crack detection procedure for structures subjected to moving vehicle
employing the particle swarm optimization method. This method can identify cracks up to
the relative crack depth of 0.1 even if in the presence of noise interference. Kang et al.
[252] have developed an improved particle swarm optimization procedure for crack
identification in beam type structure using vibration signatures. The improved method has
formulated combining the particle optimization method with the artificial immune system.
Hester and Gonzalez [253] have proposed an inventive wavelet analysis for damage
detection in bridge structure under vehicle load. The proposed method has used the
vehicle-bridge finite element interaction structure, acceleration signal and wavelet energy
content at each segment of the bridge for the structural damage isolation.
Sahoo and Maity [254] have developed a hybrid neuro-genetic mechanism for damage
isolation in different types of structure. The frequency and strain as input factors and the
damage location and severity as output factors are considered in the proposed network.
The accuracy of the network has been proved by exemplifying a fixed-free and plane
frame structure. Li et al. [255] have proposed a crack extraction method based on image
processing by placing a long distance acquisition device. The width of cracks is also
measured by implementing image clip & fill and rotation of transformation. Adhikari et
al. [256] have developed an improved model based on digital image processing for bridge
assessment. A 3-D visualization method has been formulated in such a way that the crack
can be seen as in the actual onsite visualization. Li and Hao [257] have investigated the
condition monitoring of truss bridge structure using relative displacement sensors. The
sensitivity and performance at the joint of the truss bridge structure are examined using the
displacement of ambient vibration.
2.6 Summary
From the literature review, some methods including numerical, finite element methods and
experimental analysis have been explored to study the response of the damaged structure
Chapter 2 Literature Review
31
subjected to moving mass. However, according to the author’s knowledge, the numerical
procedure, FEA and experimental verifications have not been applied simultaneously to
determine the response of deteriorated structures under moving mass with multiple cracks.
Again there are so many techniques based on FEM, experimental investigations,
computational analysis, AI- based techniques, and statistics based methods are applied for
fault isolation in deteriorated structure subjected to moving mass. But so far from the
literature review, the application of rule-based Recurrent Neural Network (Jordan and
Elman), hybridisation of rule based Jordan’s and Elman’s RNN, a rule-based statistical
properties based method and hybridisation of rule- based RNN and statistical properties
based methods are scanty for fault isolation in deteriorated structures subjected to moving
mass.
32
Chapter 3
THEORETICAL-NUMERICAL
ANALYSIS OF MULTI-CRACKED
STRUCTURES SUBJECTED TO
MOVING MASS
3.1 Introduction
The dynamic interaction between a structure and the moving mass characterizes a
particular topic in the study of structural dynamics. Theoretically, the two subsystems
(structure and moving mass), can be replicated as two elastic substructures with a general
interface. Each substructure is characterised by some frequencies of vibration. The two
substructures act together with each other through the common contact forces. In this
Chapter, a theoretical-numerical investigation of structures with different end conditions
subjected to moving mass is presented.
3.2 The Problem Description:
Here, a computational solution to the system of partial differential equation has been
developed illustrating a multi-cracked structure subjected to a traversing mass at various
end states. The computational method expresses the conversion of the well-known
principal partial differential equations of motion into a novel solution of ordinary
differential equation. A cracked cantilever structure subjected to a traversing mass is
explained in Figure 3.1. Structures with different boundary conditions are considered in
the present dissertation. The basic objectives of the problem are as follows-
(i) Formation of easy and realistic theoretical-numerical method for evaluating the
responses of multi-cracked structures subjected to transit mass with different end states.
(ii) Influences of moving mass, speed, crack depth, and crack locations and their effects on
the solution of the problem.
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
33
3.3 The Problem Formulation
Structures with different end conditions are analysed by considering the fundamental
assumptions of the Euler-Bernoulli’s beam theory. The material is assumed homogeneous
and isotropic. The rotary inertia, shearing forces, damping, and longitudinal vibrations of
the beam structure are neglected, and the transverse vibration of the beam is considered.
The inertial and shearing effects of the moving mass are accounted in the current analysis.
From the analysis of Figure 3.1, the following assumptions are made:
( )u x u = Deflection due to longitudinal vibration of the beam.
( )y x y = Deflection due to transverse vibration of the beam.
( )I x I = Moment of inertia of the beam.
( )m x m = Beam mass per unit span length (constant)
g = gravitational constant, L =Length of the beam.
B=Width of the beam, H=Thickness of the beam.
( )F t The interactive force involving the moving mass and the structure.
1 2 3 1,2,3, ,L L L L = Position of the first, second and third cracks at the left end of the
cracked structures respectively. 1 2 3 1,2,3, ,d d d d = Depth of the first, second and third
cracks respectively. M = Mass of the moving mass, v =Moving speed of the transit mass.
vt = Position of the moving mass at any instant time‘t’.
h = The point of attention where the beam deflection is to be determined.
Based on the above assumptions, the governing equation of motion of a structure under a
moving mass is given as- 4 2
4 2( , )
y yEI m F x t
x t
(3.1)
Here ( , ) ( ) ( ) ( , )F x t P t x r x t
( , )r x t Standard loading conditions (zero in the present analysis). δ=Dirac delta
function.
3.4 Analysis of Cracked Structures Subjected to Moving
Mass
The present section is focused on determining the responses of multi-cracked structures
with different end conditions subjected to a moving mass. Initially, the analysis is carried
out for a multi-cracked cantilever beam under moving mass and it is further extended to
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
34
multi-cracked simply supported beam and fixed-fixed beam under a moving mass. The
mass ‘ M ’ is moving with a speed ‘ v ’ from the fixed end of the structure to the free end as
shown in Figure 3.1. The cracks with arbitrary crack depth are located from the clamped
end of the beam.
Based on the above postulations, the equation of motion of the structure under a transit
mass is expressed as 4 2
4 2( ) ( )
y yEI m P t x
x t
(3.2)
Here
2
( , )( ) v y tt
P t Mg M
Employing the integral properties of function of Dirac delta to the structure, it may be
expressed as-
Figure 3.1: Multi- cracked cantilever beam under moving mass
M
1d
2d
3d
v
1L
2L
3L
L
u
y
( )F t
( ), ( )I x m x
x
h
H
1,2,3d
B
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
35
0 0 0
0
( ) 0 ( ) ( )
L
f x xf x x x dx L (3.3a)
0 0 0
0
( ) ( ) 0, 0,
L
x xf x x x dx L (3.3b)
Substituting the value of ( )P t in equation (3.2), one can express as-
2
2
24
4
( , ) ( )
( , )( , )
y x tEI m Mg M x
t
y x tv y t
x t
(3.4)
The universal elucidation of equation (3.4) may be written as in series from i.e.
1
( , ) ( ) ( )n n
n
y x t x Q t
(3.5)
Where ( )n x = Eigen function of the beam without considering the moving mass.
( )nQ t = Amplitude function to be evaluated, n = Number modes of vibrations.
To evaluate ( )x , the equation (3.5) can be expressed as-
4( ) ( ) 0 iv
n n nx x (3.6)
Where 2 2
4 n n
nA
EI EIm
Due to the occurrence of three numbers of cracks in the structure, the complete structure
can be replicated by the combination of four structural segments, and each segment is
obeying the adopted assumptions. In view of the beam theory of Euler-Bernoulli, the
universal solution of equation (3.6) for determining the transverse deflections for each
segment of the structure may be expressed as:
1 1 1 1 1( ) sin( ) cos( ) sinh( ) cosh( ) , 0 n n n n n
x A x B x C x D x x L (3-7a)
2 2 2 2 1 2( ) sin( ) cos( ) sinh( ) cosh( ) , n n n n nx A x B x C x D x L x L (3-7b)
3 3 3 3 2 3( ) sin( ) cos( ) sinh( ) cosh ( ) , n n n n nx A x B x C x D x L x L (3-7c)
4 4 4 4 3 ( ) sin ( ) cos( ) sinh( ) cosh ( ) , n n n n nx A x B x C x D x L x L (3-7d)
Where , , &A B C D are constant coefficients of integration, computed from different
boundary conditions of the structures [Thomson, 258]. The modelling of cracks are
explained in Appendix-A. The natures of the cracks are open and transverse.
The different boundary conditions of the cantilever beam are as follows-
At 0, (0, ) 0, (0, ) 0x y t y t and at , ( , ) 0x L L t , ϐ(L, t) =0
Where ϐ is the sheer force and is the bending moment.
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
36
Substituting the equation (3.5) on the right of the equation (3.4) and arranging it, the
equation now can be written as-
2
1 1
( ) ( ) ( ) ( ) ( )n n n n
n n
Mg M v Q t x x tt
(3.8)
With multiplication of ( )P x in the equation (3.8) and integration of the equation upon the
entire length of the beam, the equation now can be expressed as:
2
10 0 0
( )( ) ( ) ( ) ( ) ( ) ( ) ( )
L L L
p p n n p n n
n
Mg M x dxx dx x v Q t x x t dxt
(3.9)
The term on the right of the equation (3.9) can be formulated as-
1 0
( ) ( ) ( )
L
n p n
n
dxt x x
(3.10)
From the orthogonality theory and orthogonal properties, the above term can be expressed
as-
1 1 2 2
0 0 0
0
( ) ( ) ( ) ( ) ( ) ( ) ... ( ) ( ) ( ) ...
= ( ) ( ) ( ) (Vanishing the lower terms)= ( ) ( )
L L L
p p p p n
L
p p n p p p p
t x x dx t x x dx t x x dx
t x x dx t S S t
(3.11)
It is due to the reason that 0
0, ( ) ( )
,
L
p n
p
n px x dx
S n p
Recalling the theory of orthogonality, orthogonal properties and Dirac-delta function’s
integral properties, and the equation (3.9) may be expressed as-
2
1
( ) ( ) ( ) ( ) ( )p n n p p p
n
Mg M v Q t t St
(3.12)
2
1
( ) ( ) ( ) ( )p n n p
np
Mt g v Q t
S t
(3.13)
The equations (3.4) and (3.8) may be united as-
4 2
4 21
( , ) ( , )( ) ( )n n
n
EIy x t y x t
m x tx t
(3.14)
Uniting equation (3.13) and (3.14) with ‘q’ number of stages, one can express now-
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
37
24 2
4 21 1
( , ) ( , )( ) ( ) ( ) ( )n q q n
n qn
my x t y x t M
EI x g v Q tx t S t
(3.15)
Substituting equation (3.5) in (3.15), the equation may be written now-
4 2
1 1
4 2
2
1 1
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
n n n n
n n
n q q n
n qn
EI
x Q t x Q t
mx t
Mx g v Q t
S t
(3.16)
Substituting the values from the equation (3.6) in (3.16) and rearranging the equation-
2
4
,
1 1
( ) ( ) ( ) ( ) ( ) ( ) 0n n n n tt q q n
n qn
Mx EI Q t mQ t g v Q t
S t
(3.17)
The equation (3.17) have to satisfy for each values of ‘ x ’ i.e.
2
4
,
1
( ) ( ) ( ) ( ) 0n n n tt q q n
qn
EIM
Q t mQ t g v Q tS t
(3.18)
The equation (3.18) is valid to evaluate the responses of any types of structures subjected
to moving mass based on the above assumptions. The value of ‘ ( )nQ t ’ has been found out
by solving equation (3.18). Runge-Kutta fourth order technique has been employed to
explain the above equation in MATLAB domain.
3.5 Numerical Formulation of Cracked Cantilever Beam
under a Moving Mass
Numerical examples are formulated for the analysis of the proposed theory. A mild steel
cracked cantilever structure under a traversing mass has been considered with the
following dimensions.
L=100cm, B=3.9cm, H=0.5cm. M =1 kg and 2 kg, 1,2,31,2,3
dH
= Relative crack
depth=0.6, 0.25, 0.45 and 0.3, 0.55, 0.4. v = 438cm/s and 573 cm/s. Relative crack
location. 1,2,31,2,3
L
L =Relative crack Positions of first, second and third cracks from the
fixed end respectively=0.25, 0.45, 0.65 and 0.5, 0.65, 0.85.
The deflections of the cracked cantilever beam are calculated at different positions of the
moving mass along with the free end and shown in the following Figures for analysis.
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
38
Figure 3.2: Graph for Deflection of beam vs. Travelling time for
undamaged beam for 438 /v cm s
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
6
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.3: Graph for Deflection of beam vs. Travelling time
for undamaged beam for 573 /v cm s
0 0.03 0.06 0.09 0.12 0.15 0.18-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
39
Figure 3.4: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
6
7
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.5: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18-1
0
1
2
3
4
5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
40
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
6
7
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.6: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.5,0.65,0.85v cm s
0 0.03 0.06 0.09 0.12 0.15 0.18-1
0
1
2
3
4
5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.7 Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.6,0.25,0.45. 0.5,0.65,0.85v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
41
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
6
7
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.8 Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.3,0.55,0.4. 0.25,0.45,0.65v cm s
0 0.04 0.08 0.12 0.16 0.2-1
0
1
2
3
4
5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.9 Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.25,0.45,0.65v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
42
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
6
7
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.10: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s
0 0.03 0.06 0.09 0.12 0.15 0.18-1
0
1
2
3
4
5
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L
M=2kg, x=vt
M=2 kg, x=L
Figure 3.11: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
43
00.05
0.1
0.150.2
1
2
3
40
2
4
6
8
Time 't' in secMoving mass in 'kg'
Bea
m d
efle
ctio
n i
n 'c
m'(
x =
vt)
Figure 3.12: 3-D Graph for Deflection of beam vs. mass vs. Travelling time
for 1,2,3 1,2,3573 / , 0.3,0.55,0.4. 0.5,0.65,0.85v cm s
00.05
0.10.15
0.20.25
0
2
4
6-2
0
2
4
6
8
10
Time 't' in secMass in 'kg'
Def
lect
ion
of
bea
m i
n 'c
m'(
x =
L)
Figure 3.13: 3-D Graph for Deflection of beam vs. mass vs. Travelling time
for 1,2,3 1,2,3438 / , 0.6,0.25,0.45. 0.25,0.45,0.65v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
44
Figure 3.14: 3-D Graph for Deflection of beam vs. speed vs. Travelling time
for 1,2,3 1,2,32 , 0.6,0.25,0.45. 0.25,0.45,0.65M kg
00.05
0.10.15
0.20.25
0.30.35
300
400
500
600
7000
2
4
6
8
10
Time 't' in secSpeed in 'cm/s'
Beam
defl
ecti
on
in
'cm
' (x
=v
t)
Figure 3.15: 3-D Graph for Deflection of beam vs. Travelling time vs. Position
of mass for 1,2,3 1,2,32 , 573 / , 0.6,0.25,0.45. 0.25,0.45,0.65M kg v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
45
3.6 Theoretical-Numerical Solution of Cracked Simply
Supported Structure under a Moving Mass
A mass ‘ M ’is moving with a speed ‘ v ’ on a multi-cracked simply supported beam from
the supported left end to the supported right end as in Figure 3.16. The cracks are located
from the left support end of the structure. The responses of the cracked simply supported
under a moving mass are calculated using equation (3.18).The various end states of the
simply supported structure are as follows:
At 0, (0, ) 0, (0, ) 0x y t t and at , ( , ) 0, ( , ) 0x L y L t L t . (3.19)
A mild steel multi-cracked simply supported structure subjected to a traversing mass has
been considered for the numerical analysis of the following dimensions.
L= 140cm, B=4.9cm, H=0.5cm. M =1 kg and 2 kg. 1,2,3 0.2, 0.3, 0.4 and 0.35, 0.45,
0.55. v = 438 cm/s and 573 cm/s.
1,2,3 0.2857, 0.5, 0.7143 and 0.1786, 0.3571, 0.5714=Relative crack positions from the
left support end. The deflections of the cracked simply supported beam are calculated at
different positions of the moving mass along with the mid-span and analyzed in the
following Figures.
h
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
46
Figure 3.17: Graph for Deflection of beam vs. Travelling
time for undamaged beam for 438 /v cm s
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.18: Graph for Deflection of beam vs. Travelling time for
undamaged beam for 573 /v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
47
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.19: Graph for Deflection of beam vs. Travelling time
for 1,2,3 1,2,3438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.20: Graph for Deflection of beam vs. Travelling time
for 1,2,3 1,2,3573 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
48
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.21 Graph for Deflection of beam vs. Travelling time
for 1,2,3 1,2,3438 / , 0.35,0.45,0.55. 0.2857,0.5,0.7143.v cm s
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.22 Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.35,0.45,0.55. 0.2857,0.5,0.7143.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
49
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.23 Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.2,0.3,0.4. 0.1786,0.3571,0.5714.v cm s
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.24: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.2,0.3,0.4. 0.1786,0.3571,0.5714.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
50
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.25: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3438 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.26: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3573 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
51
Figure 3.28: 3-D Graph for Deflection of beam vs. speed. Travelling time
1,2,3 1,2,3 2 , 0.35,0.45,0.55. 0.1786,0.3571,0.5714M kg
00.1
0.20.3
0.4
300400
500600
700800-0.5
0
0.5
1
1.5
Time 't' in secSpeed in 'cm/s'
Defl
ecti
on o
f beam
in 'cm
' (x
=vt)
00.05
0.10.15
0.20.25
1
2
3
4
5-1
0
1
2
3
4
Time 't' in secMass in 'kg'
Defl
ecti
on
of
beam
in
'cm
'(x
= v
t)
Figure 3.27: 3-D Graph for Deflection of beam vs. mass vs. travelling time
for 1,2,3 1,2,3 573 / , 0.35,0.45,0.55. 0.1786,0.3571,0.5714v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
52
00.1
0.20.3
0.4
1
2
3
4
5-2
-1
0
1
2
3
Time 't' in secMass in 'kg'
Defl
ecti
on
of
beam
in
'cm
'(x
= L
/2)
Figure 3.29: 3-D Graph for Deflection of beam vs. mass. Travelling time
1,2,3 1,2,3 438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143v cm s
Figure 3.30: 3-D Graph for Deflection of beam vs. Travelling time vs. Position of
mass for 1,2,3 1,2,3 1 , 438 / , 0.2,0.3,0.4. 0.2857,0.5,0.7143M kg v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
53
3.7 Theoretical-Numerical Solution of Cracked Fixed-Fixed
Beam under a Moving Mass
A fixed-fixed structure with multiple cracks under a traversing mass ‘ M ’with speed ‘ v ’
is analysed as in Figure 3.31. The cracks are positioned from the fixed left part of the
structure. The responses of the multi-cracked fixed-fixed structure subjected to a transit
mass are determined by employing equation (3.18) with proper end conditions. For the
numerical analysis, a multi-cracked fixed-fixed beam of mild steel has been considered
with the prescribed dimensions.
L=140cm, B=4.9cm, H=0.5cm. M =1kg and 2kg. v = 512 cm/s and 617 cm/s. 1,2,3 0.2,
0.35, 0.45 and 0.3, 0.5, 0.55. 1,2,3 0.1429, 0.3214, 0.5357 and 0.25, 0.4286,
0.7143=Relative crack positions from the left fixed end.
1L
2L
x
h
3L
L
B
H
1,2,3d
1d
2d
3d
M v
Figure.3.31: Multi-cracked fixed-fixed beam subjected to moving mass
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
54
The deflections of the cracked fixed-fixed beam are determined at different positions of
the transit mass along with the mid-section of the beam and analyzed in the following
Figures.
0 0.05 0.1 0.15 0.2 0.25-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.33: Graph for Deflection of beam vs. Travelling time
for undamaged beam 617 /v cm s
Figure 3.32: Graph for Deflection of beam vs. Travelling time
for undamaged beam for 512 /v cm s
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
55
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Time 't' insec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.34: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357v cm s
0 0.05 0.1 0.15 0.2 0.25-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.35: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3617 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
56
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.36: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3512 / , 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s
0 0.05 0.1 0.15 0.2 0.25-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.37: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3617 / , 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
57
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.38: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3512 / , 0.3,0.5,0.55. 0.1429,0.3214,0.5357v cm s
0 0.05 0.1 0.15 0.2 0.25-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.39: Graph for Deflection of beam vs. Travelling time for
1,2,3 1,2,3617 / , 0.3,0.5,0.55. 0.1429,0.3214,0.5357v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
58
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.40: Graph for Deflection of beam vs. Travelling time
for 1,2,3 1,2,3512 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143v cm s
0 0.05 0.1 0.15 0.2 0.25-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
M=1kg, x=vt
M=1kg, x=L/2
M=2kg, x=vt
M=2kg, x=L/2
Figure 3.41: Graph for Deflection of beam vs. Travelling time
for 1,2,3 1,2,3617 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
59
Figure 3.43: 3-D Graph for Deflection of beam vs. speed vs. Travelling
time for 1,2,3 1,2,3 2 , 0.3,0.5,0.55. 0.25,0.4286,0.7143.M kg
00.05
0.10.15
0.20.25
0.30.35
500
600700
800
9001000
-1
-0.5
0
0.5
1
Time 't' in secSpeed in 'cm/s'
Defl
ecti
on o
f beam
in 'cm
' (x
=L
/2)
00.05
0.10.15
0.20.25
1
2
3
4
5-2
-1
0
1
2
Time 't' in secMass in 'kg'
Def
lect
ion
of
bea
m i
n 'c
m' (
x =
L/2
)
Figure 3.42: 3-D Graph for Deflection of beam vs. mass vs. Travelling
time for 1,2,3 1,2,3 617 / , 0.3,0.5,0.55. 0.25,0.4286,0.7143.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
60
00.1
0.20.3
0.4
1
2
3
4
5-0.5
0
0.5
1
1.5
Time 't' in secMass in 'kg'
Def
lect
ion
of
bea
m i
n 'c
m'(
x =
vt)
Figure 3.44: 3-D Graph for Deflection of beam vs. mass vs. Travelling
Time for 1,2,3 1,2,3 512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357.v cm s
Figure 3.45: 3-D Graph for Deflection of beam vs. Travelling time vs. Position of
mass for 2 ,M kg 1,2,3 1,2,3 512 / , 0.2,0.35,0.45. 0.1429,0.3214,0.5357.v cm s
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
61
3.8 Identification of cracks form the measured dynamic
response of structures
The graphs between the relative distance and relative deflections of structures under the
transit mass are shown in Figures 3.46 (Cracked cantilever beam), 3.47 (Cracked simply
supported beam) and 3.48 (Cracked fixed-fixed beam). The magnified views of cracks for
the case of cantilever beam are presented in Figures 3.49(a), (b), (c) at different relative
crack locations. It has been observed that abrupt changes in relative deflections occurred in
the dynamic response of structures. The sudden changes in relative deflections of structures
will predict the existence, locations and severities of cracks.
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
1.2
Relative distance from fixed end
Rel
ativ
e d
efle
ctio
n
Location of first crack
Location of second crack
Location of third crack
Figure 3.46: Detection of cracks for cantilever beam for 1,2,3 0.5,0.65,0.85
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
62
0 0.2 0.4 0.6 0.8 1-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Relative distance from left support end
Rel
ativ
e d
efle
ctio
n
Location of first crack
Location of second crack
Location of third crack
Figure 3.47: Detection of cracks for simply supported beam for
1,2,3 0.2857,0.5,0.7143.
0 0.2 0.4 0.6 0.8 1-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Relative distance from left fixed end
Rel
ativ
e d
efle
ctio
n
Location of first crack
Location of second crack
Location of third crack
Figure 3.48: Detection of cracks for fixed-fixed beam for
1,2,3 0.25,0.4286,0.7143.
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
63
0.04
0.06
0.08
0.1
0.12
0.485 0.49 0.495 0.5 0.505 0.51 0.515
Rel
ati
ve
def
lect
ion
Relative distance from fixed end
Undamaged
Damaged
Figure 3.49(a): Magnified view of crack for 0.5
Figure 3.49(b): Magnified view of crack for 0.65
0.17
0.19
0.21
0.23
0.25
0.27
0.29
0.635 0.64 0.645 0.65 0.655 0.66 0.665
Rel
ati
ve
def
lect
ion
Relative distance from fixed end
Undamaged
Damaged
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
64
3.9 Comparison of Results of Theoretical-Numerical and
Experimental analysis for the Response of Structures
For the validation and accuracy of the applied numerical method (Runge-Kutta method),
experimental verifications are carried out in the laboratory. Detailed analyses of the
experimental verifications have been elaborated in Chapter-8. The experiment has been
conducted with different types of structures (cantilever, simply supported and fixed-fixed
beam) under moving mass with some number of observations. For the laboratory tests, the
same configurations as those of numerical models are considered with the same beam
specimen and dimensions. The deflections of the beam are measured at different positions
of the moving mass and the specified location of the beams during the movement of the
traversed mass on the beam. The results obtained from the computational analysis, and
laboratory tests are shown in the below Tables for comparison studies. The variations of
results, time (sec) ~deflection (cm) for different structures under transit mass, obtained
from both the computational analyses and laboratory investigation are given in Tables 3.1
(Cracked cantilever beam), 3.2 (Cracked simply supported beam) and 3.3 (Cracked fixed-
fixed beam) for the comparison of results. It has been observed that the variation of results
of the numerical analysis and experimental verification are within the estimated error of
5% approximately for all the structures. So the applied numerical method (Runge-Kutta
method) yields well with the experimental verifications.
Figure 3.49(c): Magnified view of crack for 0.85
0.52
0.54
0.56
0.58
0.6
0.62
0.64
0.66
0.835 0.84 0.845 0.85 0.855 0.86 0.865
Rel
ati
ve
def
lect
ion
Relative distance from fixed end
Undamaged
Damaged
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
65
The percentage of deviation between the experimental and numerical values is given by
the relation= (Experimental Numerical)
100Experimental
Average percentage of error= Sum of the percentage errors
Total number of observations
Total percentage of error = Sum of the average percentage errors
Total number of average percentage of errors
Table 3.1: Comparison of results for beam deflection (cm) between experiment and numerical for
cracked cantilever beam for 1,2,3 1,2,3438 / . 0.6,0.25,0.45. 0.25,0.45,0.65.v cm s
Time
(sec)
Numerical
(x=vt)
Numerical
(x=L)
Experiment
(x=vt)
Experiment
(x=L)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.0342 0.0061 0.0118 -0.0061 -0.0115 0.0063 0.012 -0.0062 -0.0117
0.04 0.0101 0.0196 -0.0021 -0.0035 0.0103 0.02 -0.0023 -0.0036
0.0571 0.0287 0.0553 0.0459 0.0891 0.0295 0.0569 0.0471 0.092
0.0799 0.1055 0.1905 0.2215 0.4328 0.1089 0.1976 0.2288 0.4512
0.1027 0.2379 0.4299 0.6629 1.2206 0.2484 0.4502 0.6921 1.2787
0.1256 0.5359 0.917 1.2909 2.303 0.567 0.97 1.3641 2.4254
0.1484 1.0222 1.6768 1.9606 3.1873 1.0783 1.7849 2.0897 3.3663
0.1769 2.0423 3.2041 2.9394 4.6597 2.1698 3.4386 3.1487 4.9959
0.1988 3.0302 4.6314 3.6836 5.6087 3.2384 4.9816 3.9783 6.0595
0.2169 3.9061 5.8909 4.2145 6.3633 4.2342 6.3844 4.5929 6.9286
Average percentage of errors 4.56 4.84 4.85 4.98
Total percentage of error 4.8
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
66
Table 3.2: Comparison of results for beam deflection (cm) between experiment and numerical for
cracked simply supported beam for 1,2,3 1,2,3573 / . 0.2,0.3,0.4. 0.2857,0.5,0.7143.v cm s
Time
(sec)
Numerical
(x=vt)
Numerical
(x=L/2)
Experiment
(x=vt)
Experiment
(x=L/2)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.0428 0.0414 0.0785 0.0496 0.094 0.042
0.08 0.0505 0.0962
0.0672 0.1297 0.2383 0.1579 0.2942 0.1346 0.2454 0.1627 0.3049
0.0855 0.2458 0.4533 0.2604 0.4776 0.2552 0.4749 0.27 0.5013
0.1038 0.3538 0.6606 0.3614 0.6739 0.3731 0.6977 0.3859 0.7157
0.1222 0.4357 0.8408 0.4357 0.8408 0.4663 0.9103 0.4717 0.901
0.1466
0.4889 1.023 0.4703 0.9842 0.5302 1.117 0.5133 1.0679
0.171 0.3807 0.9143 0.4146 0.9894 0.4099 0.9815 0.4451 1.0788
0.1894 0.2632 0.7227 0.3059 0.8567 0.2798 0.763 0.3247 0.9126
0.2077 0.104 0.3622 0.1605 0.5901 0.1090 0.3751 0.1676 0.6186
0.2199 0.0242 0.1233 0.0251 0.34 0.0537 0.1265 0.0559 0.3507
Average percentage of errors 4.93 5.31 4.88 5.17
Total percentage of error 5.06
Table 3.3: Comparison of results for beam deflection (cm) between experiment and numerical for
cracked fixed-fixed beam for 1,2,3 1,2,3617 / . 0.2,0.35,0.45. 0.25,0.4286,0.7143.v cm s
Time
(sec)
Numerical
(x=vt)
Numerical
(x=L/2)
Experiment
(x=vt)
Experiment
(x=L/2)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.034 0.0113 0.0215 0.0165 0.0319 0.0115 0.0219 0.017 0.0326
0.0567 0.0541 0.1012 0.0744 0.1388 0.0554 0.1043 0.0766 0.1432
0.0794 0.12 0.2331 0.1388 0.2687 0.1248 0.2525 0.1444 0.2809
0.1021 0.1759 0.369 0.1772 0.3713 0.184 0.3899 0.1861 0.3927
0.1248 0.1892 0.4245 0.1907 0.4282 0.1997 0.4557 0.2025 0.4593
0.1418 0.169 0.395 0.1778 0.4154 0.1805 0.4295 0.1903 0.4509
0.1588 0.1172 0.2829 0.1334 0.3241 0.1269 0.3048 0.1441 0.3468
0.1759 0.0577 0.1236 0.0652 0.1515 0.0612 0.1314 0.0689 0.1599
0.1929 0.0136 -0.0075 0.0117 -0.0432 0.0144 -0.0078 0.0123 -0.0453
0.2042 0.0039 -0.006 0.0034 -0.1136 0.0041 -0.0062 0.0035 -0.1175
Average percentage of errors 5.24 4.98 4.93 4.71
Total percentage of error 4.96
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
67
3.9 Discussions and Summary
The solutions to the problem for the response of structures under transit mass have been
found out by applying Runge-Kutta fourth order rule. The solution can be implemented for
any types of the structure under transit mass with various boundary conditions. The
governing equation of motion of the dynamic mass-structural systems has been addressed
as a sequence of fixed partial differential equations. It is always assumed that there is no
separation between the moving mass and the structure. The deflections of the beam at each
location of the transit mass on the beam throughout the movement of the transit mass along
the structure and the tip end deflection (cantilever beam), and mid-span deflection (simply
supported and fixed-fixed beam) of the structure have been also calculated during the
movement of the mass across the beam at different mass, speed, crack depth and crack
locations. Numerical studies have been illustrated for various types of beam structures
subjected to transit mass for determining the deflections of the structures. From the analysis
of Figures 3.2-3.11, 3.17-3.26 and 3.31-3.41, it has been observed that the deflection
induced due to the moving mass of a damaged beam is greater than that of the undamaged
beam. Again by increasing the crack depth and weight of the traversing mass, the
deflections of the structure also increases.
In the case of a cantilever beam structure (Figures 3.2-3.11); the deflections at each location
of the moving mass ( )x vt on the beam and the tip end ( )x L of the beam are evaluated
during the movement of the transit mass on the structure. It has been observed that the tip
end deflection of the structure decreases with the enhancement of the velocity of the transit
mass. This is because, at greater velocity of the transit mass, the lower modes of the
structures are not effective and hence less deflection is observed in the structure. However
the transverse dynamic deflections of the beam mainly depend on the vibration of the lower
modes of the structure. It has also been noticed that the deflections of the beam at the tip
end reduces sharply with time and again amplify at the higher speed of the moving mass,
it’s because vibration at superior modes is more prevailing than those of inferior modes.
The vibrant deflection at the tip end of the cantilever structure depends upon the modal
excitation of the structure. So the deflection at the tip end shows sudden variation in
magnitude with direction. If the location of cracks moves towards the fixed end of the
cantilever structure, then the corresponding beam deflection also amplifies.
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
68
In the case of a simply supported beam structure, the mass is assumed to move from the left
to right. The dynamic deflections of the structure at each location of the moving mass
( )x vt on the structure and the middle ( )2
Lx of the structure are evaluated during the
passage of the transit mass along the structure. From the analysis of Figures 3.17-3.26, it
has been noticed that the dynamic deflection at the position of the moving mass ( )x vt on
the structure progressively amplifies till the mass reaches at the mid-span of the structure,
then the deflections of the structure start reducing towards the right end. But the beam mid-
span deflection ( )2
Lx initially increases till the mass reaches the mid-span, starts
reducing towards the right end. With the amplification of the traversing speed, the
deflection of the structure ( )x vt decreases up to the mid-span of the beam, then after
crossing the mid-span, it again amplifies towards the right end. However with the
amplification of speed, the deflection at the beam mid-span ( )2
Lx decreases until the
mass travels to the mid of the structure, then it starts increasing on the way to the right end.
If the crack locations approach towards the supported right end, then the displacement of
the beam decreases and increases somewhat towards the right support end.
In the case of the fixed-fixed beam, the mass is assumed to move from the fixed left end to
right end. The responses of the structure under transit mass have been analyzed in Figures
3.31-3.41. It has been noticed that the deflection of the fixed-fixed structure at the position
of the transit mass ( )x vt on the structure gradually amplifies till the traversing object
attains the middle of the structure, then the deflections start reducing towards the right end
of the structure. While the beam mid-span deflection ( )2
Lx primarily increases till the
mass traverses to middle of the beam, then it decreases gradually after the passage of the
mid-span and commence to drop quickly towards the right end of the structure. With the
enhancement of the weight of the transit mass, the deflections of the
structure ( and )2
Lx vt x gradually increase, but while approaching towards the right
end of the structure, it decreases up to some extents of the structure and then it increase
towards the end. With the amplification of the speed of the transit mass, the beam
deflection ( and )2
Lx vt x primarily decreases, and then an increase approaching till the
mass reaches at the mid-span of the beam. The beam deflections decreases while the
Chapter 3 Theoretical-Numerical Analysis of Multi-Cracked Structures
Subjected to Moving Mass
69
traversing mass approaches towards the end of the structure. If the location of cracks
approach towards the fixed right end of the structure, then the deflection produced is less.
3-D graphs are plotted and explained in Figures 3.12-3.15 (cracked cantilever beam),
Figures 3.27-3.30(cracked simply supported beam), and Figure 3.42-3.45 (cracked fixed-
fixed beam) for various masses, position of the transit mass and speed. The variation of
deflections are obtained at different positions of the transit mass on the structure ( )x vt ,
end point of the cantilever beam ( )x L , mid-span of both simply supported and fixed-
fixed beam ( )2
Lx for variation in different weights, velocities of the transit mass. The
responses of the damaged structures are also studied from the behaviour of the 3-D plots.
Similar observations are also obtained from the plot of 3-D graphs.
The laboratory tests are conducted for all the structures at different configurations of mass,
speed, relative crack depth and crack locations. The variation of results obtained from both
numerical studies and laboratory tests are illustrated in Tables 3.1,2 and 3. The results
obtained from the numerical studies agree well with experimental results and the deviation
is to be within 5%. So the applied numerical method yields well. The magnified views of
the cracks for the case of cantilever beam are shown in Figures 3.49 (a), (b) and (c). From
the measured dynamic response of the structures, the feasible existence, locations and
intensities of cracks can be predicted.
The beam displacements at any position of the transit mass on the structure have mainly
depended on the weight of the transit mass. It has been concluded that parameters like the
weight and speed of the transit mass, depth and location of cracks affect the dynamic
behaviour and response of the structure.
70
Chapter 4
FINITE ELEMENT ANALYSIS
CRACKED STRUCTURES SUBJECTED
TO MOVING MASS
4.1 Introduction
The finite element methods (FEM) with nonlinearities are frequently employed to find the
solutions to problems related to engineering structures. The competent supervision of
bridges and highway structures, where the information based on the condition of
structures, real effects of dynamic loading, the impact of speed and mass on the dynamic
characteristics of structures are various issues for preparing administration assessment and
establishing the load limit of the structures. Finite Element Analysis (FEA) can be applied
to explain the real response of structures due to moving load. The FEA of moving mass-
structure interaction dynamic was carried out using the commercial ANSYS
WORKBENCH 2015 in this Chapter. The results from the FEA are verified with those of
experiments and numerical formulations to ensure the accuracy and exactness of the
proposed analysis.
4.2 Method for FEA of moving mass-structure using
ANSYS
In ANSYS, the FEA of transit mass-structure interaction dynamic is carried out by using
transient dynamic analysis method. The transient dynamic analysis is a computational
method that is applied to evaluate the response of the structure under moving load varying
in time. This method is useful for calculating the time-varying strains, displacements and
stresses of structures. The general non-linear principal equation of motion for the transient
dynamic analysis is expressed as-
[ ] [ ] [ ] ( )t t tM x C x K x f t (4.1)
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
71
Where [ ]tM x -Inertial force, [ ]tC x -Damping force, [ ]tK x -Stiffness force, ( )f t -Applied
force. The schematic view of the vibrating system under applied force has been shown in
Figure 4.1.
To find out the solution of equation (4.1), time integration has to be carried out. The
implicit time integration method has been chosen for the present investigations. In
ANSYS, the implicit time integration technique is Newmark’s integration method. In
ANSYS Mechanical, two types of solution methods (The full method and the mode
superposition method) can be applied to find out the time-varying responses of the
structure.
The full method: This method is a simple one to set up. The entire matrices [M, C and
K] are applied to evaluate the stresses, displacements and strains in a single pass. All kinds
of nonlinearities, efficient employment of solid-model loads and regular time stepping are
allowed in this analysis. The principal drawback of this method is the requirement of more
computational time for the size of the prescribed model.
The mode superposition method: The fundamental concept of this method is to explain
the responses of a structure to the linear combination of its all undamped mode shapes.
This approach is quicker and requires less computational time than that of the ‘The full
method’. Damping is allowed as in terms of frequency. The step of time is fixed. The
disadvantage of this method is that the nonlinearities are not allowed.
The full method, transient dynamic analysis, is applied in the present analysis due to the
inclusion of nonlinearities. In this analysis, Newmark’s time integration method under
zero damping conditions has been applied to evaluate the response of structures in
K C
U
M
Figure 4.1: Free body diagram of vibrating system
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
72
ANSYS WORKBENCH 2015 domain. Newmark’s integration parameters under the
schemes of unconditionally stable and constant average acceleration method are
considered in this analysis.
4.3 Steps involving ‘The full method’ transient dynamic
analysis in ANSYS
Before analysing the different steps of transient dynamic analysis, modal analysis is to be
completed to extract the mode shapes and natural frequencies of the structures.
Step 1- Select Transient structural
Step 2- Select Material properties (material types, ρ, E, υ, k, G).
Step 3- Build the geometry
Step 4- Select Model-Contact (No separation) - Sliding is allowed
Step 5- Mesh – (Element size)
Step 6- Select Transient- Initial conditions
Step 7- Analysis setting- Give the ending time of step
-Automatic time stepping is allowed (on)
-Initial step timing = 120
-Set minimum step timing
-Set maximum step timing
-Time integration (on)
-Standard earth gravity (g)
-End conditions
-Speed
Step 8- Select solution- Directional deformation
Here ‘ ’ is the highest natural frequency of the structure.
4.4 Response analysis of cracked structures under moving
mass using ANSYS
The FEA for the dynamic response of structures subjected to transit mass has been
performed using the commercial ANSYS WORKBENCH 2015. In the primary stage,
modal analyses up to three modes of vibration are performed to extract the mode shapes
and the natural frequencies of the structures.
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
73
In ANSYS, the responses of the structures at different locations of the transit mass and the
particular location of the structure are calculated. The dimensions of the structures are
same as those of experimental model with the same damage configurations, traversing
mass and speed. The interaction of transit mass-structure dynamic for the case of the
Figure 4.2 Transit mass-structure interaction of cracked cantilever beam for
1,2,3 1,2,30.3,0.55,0.4. 0.5,0.65,0.85. M=2 kg
Figure 4.3: Magnified view of crack for α=0.55
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
74
cantilever structure is shown in Figure 4.2. The crack modelling is also carried out in
ANSYS. The magnified view of a crack is shown in Figure 4.3.
Table 4.1: Frequencies ratios of damaged cantilever beam
Mode
No
1,2,3 0.6,0.25,0.45.
1,2,3 25,45,65L cm
1,2,3 0.3,0.55,0.4.
1,2,3 50,65,85 .L cm
1,2,3 0.6,0.25,0.45.
1,2,3 50,65,85 .L cm
1,2,3 0.3,0.55,0.4.
1,2,3 25,45,65L cm
1 0.9908 0.9971 0.9674 0.9843
2 0.9587 0.9728 0.9833 0.9662
3 0.9904 0.9718 0.9638 0.9824
Length of the beam
Def
orm
atio
n o
f th
e bea
m
Figure 4.4 (a): Second mode shape of cantilever structure
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
75
Figure 4.4 (b): Third mode shape of cantilever structure
Length of the beam
Def
orm
atio
n o
f th
e b
eam
Figure 4.5: Schematic view of transient structural model for cracked cantilever beam
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
76
Table 4.2: Comparison of results for beam deflection (cm) between experiment and FEA for
cracked cantilever beam for 1,2,3 1,2,3573 / . 0.3,0.55,0.4. 0.5,0.65,0.85.v cm s
Time
(sec)
Experiment
(x=vt)
Experiment
(x=L)
FEA
(x=vt)
FEA
(x=L)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.0262 0.0056 0.0105 -0.0075 -0.0146 0.0055 0.0103 -0.0074 -0.0144
0.0436 0.0289 0.054 0.0073 0.015 0.0281 0.0529 0.0072 0.0147
0.0611 0.0687 0.1291 0.1144 0.2163 0.0665 0.1256 0.1116 0.2114
0.0785 0.1444 0.2618 0.3646 0.6966 0.1394 0.2541 0.3542 0.6788
0.096 0.3309 0.5724 0.7786 1.4428 0.3189 0.5536 0.7546 1.4009
0.1134 0.6791 1.124 1.290 2.1807 0.6535 1.0857 1.2496 2.1101
0.1309 1.176 1.8549 1.866 2.9291 1.1313 1.7894 1.8024 2.8292
0.144 1.6694 2.5523 2.3412 3.6142 1.6044 2.4581 2.2574 3.4877
0.1571 2.4058 3.5935 2.7968 4.1889 2.3115 3.4548 2.6933 4.0343
0.1658 2.8857 4.2526 3.1063 4.5708 2.7712 4.0834 2.9858 4.3903
Average percentage of errors 3.47 3.12 3.03 2.92
Total average percentage of error 3.13
The percentage of error between the experimental and FEA values are given by the
following relation, Percentage of error= (Expt.values-FEA values)100
Expt.values .
Average percentage of error= Sum of the percentage errors
Total number of observations
Total percentage of error = Sum of the average percentage errors
Total number of average percentage of errors
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
77
Table 4.3: Comparison of results for beam deflection (cm) among experiment, FEA and numerical
for cracked cantilever beam for M=1kg,v=438cm/s, 1,2,3 1,2,30.3,0.55,0.4. 0.5,0.65,0.85
Time
(sec)
Experiment FEA
Numerical
(x=vt) (x=L) (x=vt) (x=L) (x=vt) (x=L)
0.0457 0.0155 0.0073 0.0152 0.0072 0.0151 0.0071
0.0685 0.0515 0.1117 0.0502 0.1096 0.0496 0.1082
0.0913 0.1295 0.3818 0.1257 0.3727 0.124 0.3675
0.1142 0.2893 0.8606 0.2802 0.8362 0.2759 0.8213
0.137 0.6626 1.4647 0.6411 1.4171 0.6297 1.3894
0.1541 1.091 2.0048 1.0549 1.9378 1.0348 1.8962
0.1712 1.5984 2.5734 1.5441 2.4837 1.5111 2.4244
0.1884 2.2184 3.137 2.1417 3.0175 2.0929 2.9412
0.2055 3.1476 3.7036 3.0372 3.5544 2.9634 3.4606
0.2226 4.0492 4.2408 3.9023 4.0636 3.8028 3.9474
Average percentage of errors 3.11 3.07 4.91 5.02
Total average percentage of error 4.03
0 0.05 0.1 0.15 0.2 0.25-1
0
1
2
3
4
5
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
Numerical, x=vt
FEA, x=vt
Expt., x=vt
Numerical, x=L
FEA, x=L
Expt., x=L
Figure 4.6: For cracked cantilever beam for 1 , 438 / ,M kg v cm s
1,2,3 1,2,30.3,0.55,0.4. 0.25,0.65,0.85
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
78
For the cantilever structure, before analysing the transient dynamic analysis, the mode
shapes and the natural frequencies of the structure are determined with different damage
configuration of the structure by considering the first three modes of vibration. The
proportions of the natural frequencies for cantilever beam are presented in Table 4.1 with
various damage configuration of the structure. The different mode shapes of the cantilever
are shown in Figures 4.4(a) and (b). The responses of the damaged cantilever beam are
found out using the full method transient dynamic analysis. The schematic view of the
transient structural model in ANSYS domain is explained in Figure 4.5 for the cracked
cantilever beam. In ANSYS, Figure 4.5, the symbol, M(y), presents the deflections of the
structures at the locations of the transit mass, and L(y) represents deflections of the
structure at the free end. The results from the experimental verification are compared with
those of FEA and numerical for the validation of the FEA method. The comparison of
results, time (sec) ~ deflections (cm), between experiments and FEA are presented in
Table 4.2, and those of among experiments, FEA and numerical are presented in Table
4.3. It has been observed that the experimental results agree well with those of result from
FEA with an average error about 3%. So the applied integration method in ANSYS,
Newmark integration method, converges well with the experiment. So the applied method
in FEA is a valid one.
Length of the beam
Def
orm
atio
n o
f th
e bea
m
Figure 4.7(a) Second mode shape of simply supported beam
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
79
Table 4.4: Frequencies ratios of damaged simply supported beam Mode
No 1,2,3 0.2,0.3,0.4.
1,2,3 40,70,100L cm
1,2,3 0.35,0.55,0.45.
1,2,3 40,70,100L cm
1,2,3 0.2,0.3,0.4.
1,2,3 25,50,80L cm
1,2,3 0.35,0.55,0.45.
1,2,3 25,50,80L cm
1 0.9816 0.9647 0.9844 0.9622
2 0.9907 0.9795 0.9947 0.9875
3 0.9921 0.9812 0.9952 0.9866
Table 4.5: Comparison of results for beam deflection (cm) between experiment and FEA for
cracked simply supported beam for 1,2,3 1,2,3438 / . 0.35,0.45,0.55. 0.1786,0.3571,0.5714.v cm s
Time
(sec)
Experiment
(x=vt)
Experiment
(x=L/2)
FEA
(x=vt)
FEA
(x=L/2)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.0479 0.037 0.0708 0.0628 0.1204 0.0365 0.0698 0.0618 0.1188
0.0799 0.1664 0.3097 0.219 0.4067 0.1636 0.3038 0.2148 0.4001
0.1039 0.2945 0.5578 0.3641 0.6775 0.2889 0.545 0.3562 0.6641
0.1279 0.4624 0.8934 0.4752 0.9142 0.4518 0.8703 0.4635 0.8943
0.1518 0.54 1.1024 0.5428 1.105 0.5257 1.0699 0.5278 1.0767
0.1758 0.5387 1.1856 0.5489 1.2108 0.5226 1.146 0.5295 1.1747
0.2078 0.4672 1.1647 0.4577 1.1462 0.4524 1.1227 0.4399 1.1068
0.2317 0.3309 0.882 0.3541 0.9723 0.3191 0.8471 0.3411 0.9344
0.2557 0.1937 0.5035 0.2389 0.6515 0.1866 0.4885 0.2308 0.629
0.2797 0.0785 0.1280 0.1368 0.2528 0.0755 0.1246 0.1327 0.2448
Average percentage of errors 2.7 2.76 2.85 2.65
Total percentage of error 2.74
Length of the beam
Def
orm
atio
n o
f th
e b
eam
Figure 4.7(b) Third mode shape of simply supported beam
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
80
Table 4.6: Comparison of results for beam deflection (cm) among experiment, FEA and numerical
for cracked simply supported beam for M=1kg, v=573cm/s,
1,2,3 1,2,30.35,0.45,0.55. 0.1876,0.3571,0.5714.
Time
(sec)
Experiment FEA
Numerical
(x=vt) (x=L/2) (x=vt) (x=L/2) (x=vt) (x=L/2)
0.0428 0.0425 0.0585 0.0418 0.0576 0.0414 0.0569
0.0733 0.1913 0.2327 0.1876 0.2281 0.1852 0.2248
0.0977 0.3998 0.4124 0.3897 0.4035 0.3836 0.3969
0.1161 0.5378 0.5374 0.522 0.5225 0.5124 0.5129
0.1344 0.6226 0.6263 0.6028 0.6067 0.589 0.5935
0.1527 0.6889 0.659 0.6643 0.6357 0.6479 0.6199
0.171 0.5948 0.6259 0.5719 0.6019 0.5567 0.5852
0.1894 0.4125 0.5053 0.3952 0.4898 0.3841 0.4755
0.2077 0.1759 0.3069 0.1707 0.298 0.1672 0.2919
0.226 0.0157 0.0483 0.0154 0.0471 0.0152 0.0463
Average percentage of errors 2.85 2.74 4.8 4.72
Total percentage of error 3.77
Table 4.7: Frequencies ratios of damaged fixed-fixed beam Mode
No 1,2,3 0.3,0.5,0.55.
1,2,3 20,45,75L cm
1,2,3 0.2,0.35,0.45.
1,2,3 35,60,100L cm
1,2,3 0.3,0.5,0.55.
1,2,3 35,60,100L cm
1,2,3 0.2,0.35,0.45.
1,2,3 20,45,75L cm
1 0.9834 0.9949 0.9893 0.9898
2 0.9852 0.9878 0.9794 0.9942
3 0.9815 0.9902 0.9817 0.9881
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
81
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time 't' in sec
Defl
ecti
on
of
beam
in
'cm
'
Numerical, x=vt
FEA, x=vt
Expt., x=vt
Numerical, x=L/2
FEA, x=L/2
Expt., x=L/2
Figure 4.8: For cracked simply supported beam for 2 , 438 /M kg v cm s ,
1,2,3 1,2,30.35,0.45,0.55. 0.2857,0.5,0.7143 .
Length of the beam
Def
orm
atio
n o
f th
e bea
m
Figure 4.9(a) Second mode shape of fixed-fixed beam
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
82
Table 4.8: Comparison of results for beam deflection (cm) between experiment and FEA for
cracked fixed-fixed beam for 1,2,3 1,2,3512 / . 0.3,0.5,0.55. 0.25,0.4286,0.7143.v cm s
Time
(sec)
Experiment
(x=vt)
Experiment
(x=L/2)
FEA
(x=vt)
FEA
(x=L/2)
M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg M=1 kg M=2 kg
0.0547 0.0325 0.0622 0.0544 0.1043 0.032 0.061 0.0536 0.1025
0.082 0.0955 0.1877 0.1177 0.2327 0.0936 0.1836 0.1154 0.2271
0.1094 0.1517 0.3136 0.1757 0.3636 0.1482 0.3046 0.1714 0.353
0.1299 0.2034 0.4308 0.203 0.4338 0.1975 0.4168 0.1969 0.419
0.1504 0.2271 0.4804 0.2264 0.4866 0.2198 0.4629 0.2192 0.4687
0.1709 0.2096 0.4553 0.2151 0.4815 0.2019 0.4373 0.2077 0.4627
0.1982 0.1185 0.2771 0.1027 0.3026 0.1149 0.2685 0.0988 0.2933
0.2188 0.0633 0.0672 0.0585 0.0739 0.0616 0.0655 0.0567 0.0721
0.2324 0.0258 0.0053 0.0334 -0.0568 0.0252 0.051 0.0325 -0.0554
0.2461 0.0097 0.0104 0.038 -0.0325 0.0094 0.0102 0.0372 -0.0319
Average percentage of errors 2.65 2.77 2.72 2.79
Total percentage of error 2.73
Length of the beam
Def
orm
atio
n o
f th
e b
eam
Figure 4.9(b) Third mode shape of fixed-fixed beam
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
83
Table 4.9: Comparison of results for beam deflection (cm) among experiment, FEA and numerical
for cracked fixed-fixed beam for M=2kg,
v=617cm/s, 1,2,3 1,2,30.3,0.5,0.55. 0.25,0.4286,0.7143.
Time
(sec)
Experiment FEA
Numerical
(x=vt) (x=L/2) (x=vt) (x=L/2) (x=vt) (x=L/2)
0.0511 0.0772 0.1183 0.0762 0.1166 0.0751 0.1152
0.0681 0.1803 0.221 0.1775 0.217 0.1747 0.2137
0.0851 0.2864 0.3362 0.2808 0.3291 0.2754 0.3231
0.1078 0.4489 0.4489 0.4383 0.4379 0.4285 0.4289
0.1248 0.5043 0.5086 0.4908 0.4947 0.4783 0.4823
0.1361 0.5049 0.5235 0.489 0.5074 0.475 0.4924
0.1532 0.4227 0.4637 0.4060 0.4478 0.3936 0.4337
0.1702 0.2507 0.2943 0.2418 0.2831 0.2352 0.2739
0.1872 0.0269 0.0323 0.0261 0.0312 0.0255 0.0305
0.2156 -0.0026 -0.1106 -0.0025 -0.1072 -0.0024 -0.105
Average percentage of errors 2.66 2.72 4.87 4.96
Total percentage of error 3.8
In the case of the damaged simply supported and fixed-fixed beam under transit mass, the
similar procedure like cantilever beam has been carried out to find the response of the
0 0.05 0.1 0.15 0.2 0.25-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Time 't' in sec
Def
lect
ion
of
bea
m i
n 'c
m'
Numerical, x=vt
FEA, x=vt
Expt., x=vt
Numerical, x=L/2
FEA, x=L/2
Expt., x=L/2
Figure 4.10: For cracked fixed-fixed beam for 2 , 617 / M kg v cm s ,
1,2,3 1,2,30.2,0.35,0.45. 0.1429,0.3214,0.5357. .
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
84
structure in ANSYS domain. The frequency ratios of the structures at different damage
configuration are presented in Tables 4.4 (damaged simply supported beam) and
4.7(damaged fixed-fixed beam). The different mode shapes of the cracked structures are
shown in Figures 4.7(simply supported beam) and 4.9(fixed-fixed beam). The responses
of the structures are determined at each location of the transit mass and mid-span of the
structure at various damage scenarios of the structures. The results from the experiments
are compared with those of FEA and numerical analysis for the validation of the proposed
FEA scheme. It has been remarked that the error between the results of experimental
analysis and FEA are near about 2.7% for both the simply supported and fixed-fixed
structures.
4.5 Discussions and Summary
The FEA of the cracked structures subjected to transit mass has been carried out applying
the full method transient dynamic analysis in ANSYS WORKBENCH 2015 domain. The
computational method analysed in the full method transient dynamic analysis in ANSYS
domain is the Newmark-time integration method. Before analysing the transient dynamic
analysis, modal analyses are carried out for all the structures up to the first three modes of
vibration. The frequencies ratios of the multi-cracked structures are presented in Tables
4.1 (cracked cantilever beam), 4.4 (cracked simply supported beam), 4.7 (cracked fixed-
fixed beam) at various damaged configurations of the structures. The different mode
shapes of the damaged structures are shown in Figures 4.4(a, b) for cantilever structure,
Figures 4.7(a, b) for simply supported beam and Figures 4.9 (a, b) for the fixed-fixed
structure. In FEA, the responses of the structures have been found out at different
positions of the transit mass and the desired positions of the structure during the passage
of the transit mass across the structures. The experimental results are compared with those
of FEA in Tables 4.2(cracked cantilever beam), Table 4.6(cracked simply supported
beam) and Table 4.8 (cracked fixed-fixed beam). The comparison of results among
experiments, FEA and numerical analyses are also explained for damaged cantilever beam
(Table 4.3, Figure 4.6), simply supported beam (Table 4.6, Figure 4.8) and fixed-fixed
beam (Table 4.9, Figure 4.10). The errors between the results of numerical analyses and
FEA are about 1.9 % for cracked cantilever beam, 1.97% for cracked simply supported
beam and 2.36% for fixed-fixed beam. Regarding the response of the structures under
transit mass, similar observations are obtained in FEA like numerical and experimental
Chapter 4 Finite Element Analysis Cracked Structures Subjected to Moving Mass
85
observations. It has been observed that results from the experiments are converged well
with the FEA with error about 3% for cantilever beam structure and 2.7% for the simply
supported and fixed-fixed beam structures. So the applied numerical method in ANSYS,
(Newmark’s time integration method), is an appropriate one to study the response of any
kind of structures under time-varying load.
86
Chapter 5
APPLICATION OF RECURRENT
NEURAL NETWORKS FOR DAMAGE
IDENTIFICATION IN STRUCTURES
UNDER MOVING MASS
5.1 Introduction
The features of all the real world structures are the realities that these are susceptible to
faults, natural calamities, breakdown, and more in general, unpredicted means of
performance. So they need a continuous system for consistent and complete monitoring of
structures based on efficient and appropriate fault diagnosis approach. This is happening
for engineering structures, whose complication is rising due to the mechanized
progression, unavoidable expansion of new industry along with the growing information
and technologies. The real design and safe operation of engineering structures focus on the
consistency, accessibility, fault tolerance and safety. So, it is the usual process for fault
diagnosis of structures in up to date control theory and practice. As a result, verities of
fault diagnosis methods arise for the better condition monitoring of structures.
Applications of artificial neural networks (ANN) are widely studied for the last two
decades and employed to fault diagnosis in structure along with the problems on
modelling of the dynamic system. ANNs present an exciting and precious option to
traditional approaches because the most complex problems are not adequately described
for the execution of deterministic algorithms. ANNs provide an exceptional mathematical
tool to analyze the non-linear problems. The features of ANNs are quiet useful for solving
problems in pattern recognition and their abilities for self-learning. In this Chapter,
recurrent neural networks (RNNs) are applied for damage identification in structures
subjected to transit mass. The Jordan’s RNNs, Elman’s RNNs, and the hybridization of
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
87
the Jordan’s and Elman’s RNNs are employed for the fault detection of transit mass-
structure interaction problems.
5.2 Overview of neural networks
The neural networks can extort the features of the structure from the chronological
training data by the learning mechanism, knowing little or no previous information about
the method. This process gives the non-linear modelling of structures with greater
flexibility. The adaptive control structures are designed for unidentified, intricate and
non-linear dynamic procedures. The NNs can also work out strongly even if in the
presence of missing and incorrect data. The defensive imparting based on ANNs is also
not influenced by an alteration in the structure operating conditions. The NNs are also
capable for massive input error tolerance, significant computation rates, and adaptive
potential. Based on the training processes, the ANNs are classified into two broad
categories i.e. feed forward neural network (FFNNs) and recurrent neural networks
(RNNs).
5.2.1 Feed forward neural networks
The simplified structure of a feed forward ANN with multi-inputs and multi-outputs is
explained in Figure 5.1. Where i, j, k are number of neurons in the input, hidden and
output layers respectively.wij and wjk are synaptic weights of the input and hidden layers
respectively. f(.) is the activation function. Wi is the input values, Y1and Y2 are output
Hidden layers Input layers
Output layers wij wjk
Y1
f (.)
f (.)
f (.)
f (.)
f (.)
W1
W2
Wi
Y2
Figure 5.1: Simplified NN model with feed forward networks
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
88
values. In feed forward neural networks (FFNNs), the computation of the network is done
only in the forward direction. Synaptic weights are assigned to the inputs to the neuron
which can influence the abilities to take decision of the neural network model. The inputs
to the neuron are named as weighted inputs. The weighted inputs are gathered in the
hidden layers. If the summation of the weighted inputs goes beyond the predetermined
threshold value, the neuron electrifies, otherwise does not electrify. The scope of the
activation function is to restrict the output of the neuron amplitude.
5.2.2 Recurrent neural networks (RNNs)
The RNNs are those kinds of neural networks which have single or multiple feedback
loops. Due to the connections of feedbacks loops to the network structure, the information
can be accumulated and used later. The use of RNNs is preferred over FFNNs due to its
dynamic memory, self-recurrent and redundancy. The simple architecture of an RNN
structure is presented in Figure 5.2.
Here W, Y = Values of the input and output layers respectively, Z-1
= The delay units from
the output to the context layers, w= Synaptic weight of the neuron. i, j, k and l= The total
number of units (neurons) in the input, hidden, output and context layers respectively.
In the RNN model, the context units are formulated due to the feedback connections. Due
to the feedback connections, the structure can form a closed loop. In the context units, the
information can be collected and employed later. The context units act as an additional
memory to the network. The recurrences in the recurrent networks permit the network to
retain information from the earlier period and used it later. The recurrences in the network
Input units Hidden units Output units
Context units
Figure 5.2: Architecture of a RNN model
Z-1
W Y
wij wjk
wlj
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
89
act as a dynamic memory to the network structure. The feedbacks in the RNN can be
either of a global or local type.
Locally recurrent networks- In the locally recurrent networks, the feedbacks are
surrounded by neuron models. In this kind of network, the feedbacks are neither connected
to the neurons of progressive layers nor horizontal connections between the neurons of the
same layer. The structure of the locally recurrent neural networks is alike to static feed
forward structure, but the structure consists of dynamic memory neuron models.
Globally recurrent networks- In these types of networks, the feedbacks are either
connected between neurons of different layers or same layer. These networks include a
static multilayer perceptron. The networks also allow the non-linear mapping potentials of
the multilayer perceptron. The globally recurrent networks are classified into three types
i.e. fully recurrent and partially recurrent networks. The partially recurrent networks have
advantages over the fully recurrent networks that these recurrent connections are well
structured. These networks lead to greater training processes and less stability. The
numbers of units are strongly linked to the number of hidden or output neurons which
effectively control their flexibility. So additional recurrent links named as context units are
provided from the hidden or output units. Based on the additional units, the recurrent
networks are broadly classified into three types i.e. Jordan’s recurrent neural networks
(JRNNs), Elman’s recurrent neural networks (ERNNs), and Hopfield’s recurrent neural
networks (HRNNs).
Jordan’s recurrent neural networks (JRNNs)-
The simple architecture of a JRNNs model is presented in Figure 5.3. The architecture of
JRNNs was developed by Michel I. Jordan. Jordan [230] has added recurrent connections
from the network’s output to form context units. In Jordan networks, the context units are
also consigned to as the position layer. The outputs linked with every state are fed back to
the context units and merged with the inputs by characterizing the next state on the nodes
of the input. The entire state now comprises a new step for progression at the subsequently
time step. After numerous steps of procedures, the model presents on the context units,
along with the input units, is representative of the particular progression of the state that
the network has performed.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
90
Elman’s recurrent neural networks (ERNNs)-
The ERNN is a partial recurrent neural network that recognizes patterns from the series of
values employing the back propagation method through time learning mechanism. Elman
[230] first developed the architecture in 1990. The simple architecture of an ERNN model
is shown in Figure 5.4. In ERNN, the recurrent links are given from the hidden layer to
the addition layer, which is known as context layer. The information from the hidden layer
is stored in the additional or context layer and again accumulated to the hidden layer. The
values from the preceding steps can be gathered and reused in the present time steps. The
Figure 5.3: Simple Architecture of JRNN model
Input layers Hidden layers Output layers
Context layers
Z-1
Z-1
Y1
Yn
W1
Wn
wij wjk
wlj
Figure 5.4: Simple architecture of ERNN model
Input layers Hidden layers Output layers
Context layers
Z-1
Z-1
Y1
Y2
W1
W2
wij wjk
wlj
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
91
accumulated information can be utilized later and this allows sequential and spatial pattern
recognition mechanism for the ERNN model.
Hopfield’s recurrent neural networks (HRNNs)-
A Hopfield is one type of RNN which was developed by Hopfield in 1982 [265]. The
model of simple HRNNs is shown in Figure 5.5. The HRNNs provide a context-
addressable memory structures with bidirectional threshold outputs. The HRNNs store
information both from psychology and neurology and develop a human memory model
that is known as associative memory. The HRNNs is a neural network which has
recognition of associative memory. The HRNNs are guaranteed to congregate a local
minimum but sometimes a wrong one. The HRNN structure provides an NN model for
recognizing the human memory.
The present Chapter is focused on knowledge-based studies on Jordan’s, Elman’s, and the
hybrid structure of both the Jordan’s and Elman’s recurrent neural networks. The results
obtained from the analyses of all network models are also compared with each other. The
knowledge-based studies are usually based on proficient and qualitative analysis. These
studies comprise the rule-based method. The rule-based methods, where the investigative
w11
w12 w13
w14 w21
w22 w23
w24 w31
w32 w33
w34
w41
w42
w43
w44
Y1 Y2 Y3 Y4
W1 W2 W3 W4
Figure 5.5: Simple architecture of HRNN model
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
92
rules can be invented from the structural process, unit function and simulation based
qualitative approach. In the present analysis, the faults are frequently identified by
reasoning some training symptoms with forward and backward directions along with the
propagating path of the neurons.
5.3 Use of Levenberg-Marquardt back propagation method
for RNN
The Levenberg-Marquardt (L.M) was developed by K. Levenberg and D. Marquardt [265]
independently. The advantages of the L.M back propagation algorithm over other
algorithms are that it is stable and fast. The L.M algorithm presents a computational
solution to a problem to minimize the non-linear function. This method is fast and has
steady convergence. The L.M algorithm is, in fact, the blend of two minimization process
i.e. the steepest descent method and Gauss-Newton method. The method comes from the
speed improvement of the Gauss-Newton technique and steadiness of steepest descent
technique. The fundamental concept of the L.M mechanism is that it executes an
interactive training process around the region with complex curvature. The L.M algorithm
changes to the steepest descent method to formulate a quadratic approximation and then
turned into the Gauss-Newton method to accelerate the convergence of the algorithm
during the training process.
According to, Yu and Wilamoski [266], the equation of the Levenberg-Marquardt back
propagation algorithm is given by-
1
1 ( ) T
k k k k k kw w J J X J e (5.1)
Where ‘ J ’ (the Jacobian matrix) has been calculated from the Gauss-Newton method. ‘X’
is the identity matrix. ‘ ’ is the combination coefficient. If the value of ‘ ’ approximates
to zero, the equation (5.1) will behave like Gauss-Newton method and, if the value of ‘ ’
is very large, then it will proceed as steepest descent method.
= (1/ ν), ‘ν’ is the step size or training constant.
e=Error vector= desired actual .
Where desired is the required output vector, actual is the actual output vector.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
93
ε= Error function= 2
1
2 all training all outputspatterns
e (5.2)
The implementation of the L.M algorithm mainly depends on the calculation of ‘J’ and
organization of the training procedure iteratively for weight updating. In the L.M back
propagation method, the back propagation process is repeated for each output so as to
attain the successive rows of the Jacobin matrix. The error back propagating units is also
determined for each neuron (hidden and output) separately in L.M algorithm both for
forward and backward computation. Once the calculation of the Jacobian matrix is over,
then the next step is to arrange the training procedure of the network.
5.3.1 Steps for the organization of the training procedure using L.M
algorithm
Step 1- Generate the initial weight (wk)
Step 2- Determine the error (εk)
Step 3- Calculate the Jacobian matrix (J)
Step 4- As per equation (5.1); update the values to adjust the weight
1
1 ( ) T
k k k k k kw w J J X J e
Step 5- Calculate the total error (εk+1) with the modified weight
Step 6- If εk+1 > εk, Eliminate the step and improve the value of ‘ ’ with some factor and
repeat step-4 to update the value.
Step 7- If εk+1 ≤ εk, Allow the step and decrease the value of ‘ ’ with some factor and
repeat step as in step-5.
Step 8- Go to step 4 with the modified weight till calculated error is lesser than the
requisite error.
Like these procedures, the L.M algorithm is updated. The Jacobian matrix is computed,
and training processes are designed. As per the update rule, if the computed error becomes
smaller than the last error, it means that the quadratic approximation on the total error is
functioning, and the value of ‘ ’ should be decreased to minimize the significance of
gradient descent section. On the other hand, if the computed error is greater than the last
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
94
error, it is required to estimate a requisite curvature for proper quadratic approximation
and the value of ‘ ’ is to be increased.
5.4 Application of rule-based modified JRNNs for damage
identification in structure under moving mass
In the present analysis, rule-based JRNNs have been proposed for damage identification in
structures under moving mass and the architecture of the developed model has been shown
in Figure 5.6. The network consists of one input, output and context layers, and three
hidden layers. There is a main feedback connection from the output layer to the context
layer. The proposed network architecture is slightly modified to the original JRNN
network by introducing self-recurrent links in the output and context units. Thus, the
dynamic memory is provided employing feedback and self-recurrent links to the network.
The nodes in the context layer receive information through the feedback links from the
output layer, and outputs of nodes in the context layer provide information to the initial
hidden layer. Each node in the context, as well as, output layers has self-recurrent links to
itself. The self recurrent and the feedback links have one-time delay unit. The numbers of
units in the context unit are same as those of output units. It’s because the output values
can be exactly copied to the context units due to the feedback links. All the feed forward,
feedback and self-recurrent links of the proposed network are adapted by the Levenberg-
Marquardt algorithm which has been presented in equation (5.1). The proposed JRNN
model for damage identification in structure has been trained with 600 patterns of data for
attributing different conditions of the each of the structural system.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
95
The adopted activation function for the hidden and context layers is ‘tan-sigmoid’ and that
of output the layer is ‘purelin’ functions respectively. The symbols used for the activation
functions in the hidden and output layers are f (.) and g (.) respectively. The self-recurrent
links also have one-time unit delay. During the training procedures and operation of the
network, the training patterns fed forward to the network encompass the following
components:
Figure 5.6: Architecture of modified JRNN model
Input layer
First hidden
layer
Output layer
Context layer
RD-1
v (m/s)
M (kg)
rcl1
rcd1
rcl2
rcd2
rcl3
rcd3
Z-1
Third hidden
layer
Second hidden
layer
(15 neurons) (15 neurons) (15 neurons)
(6 neurons)
(6 neurons)
(6 neurons)
RD-3
RD-2
RD-4
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
96
t=Total travelling time of the traversing mass to cross the beam
RD= Relative deflection= Deflection of cracked beam to uncracked beam at a particular
instant of time.
RD-1 =Relative deflections of the beam at time‘t/4’.
RD-2 = Relative deflections of the beam at time‘t/2’
RD-3= Relative deflections of the beam at time ‘3t/4’.
RD-4= Relative deflections of the beam at time‘t’.
W1= RD-1. W2= RD-2. W3= RD-3. W4= RD-4
W5= Speed of the moving mass (v).
W6= Magnitude of the moving mass (M).
1 = Relative first crack location (rcl1).
2 = Relative first crack depth (rcd1).
3 = Relative second crack location (rcl2).
4 = Relative second crack depth (rcd2).
5 = Relative third crack location (rcl3).
6 = Relative third crack depth (rcd3).
Where, i= 1, 2...N1, ‘N1’ is the total input nodes. r= 1, 2...N2, ‘N2’ is the total context
nodes. k=1, 2...O, ‘O’ is the total output nodes.
‘j1= j2 = j3=1, 2... S’, ‘S’ is the total number of neurons (nodes) in each of the first, second
and third hidden layers respectively (constant for all the nodes).
‘β’ is the self-recurrent value of the each node in the context and output layers respectively
which lie between 0 to 1.
Z-1
is the unit delay.
U1-6= The values of context units in the context layer.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
97
1t
r
and t
r are the output values of the context node ‘r’ at time index ‘t-1’ and ‘t’
respectively.
t
j is the output values of the hidden node ‘j’ at time index ‘t’.
1t
k
and t
k are the output values of the output nodes at time index ‘t-1’ and ‘t’
respectively.
‘t-1’ is the time index which is delayed by one-time step due to the feedback links.
‘w’ is the weight of connection.
From the analysis of the network model (Figure 5.1), it has been observed that
1 1t t t
r k r (5.3)
The net input to the first hidden layer, 1 , 1 , 1
1 1
t tN Rt
j i i j r r j
i r
W w w
(5.4)
The net input to the second hidden layer, 2 1 1, 2
1
tSt
j j j j
j s
w
(5.5)
The net input to the third hidden layer or network is given by the following relation-
3 2 2, 3
2 1
tSt t
j j j j j
j
net w
(5.6)
( )t t
j jf net (5.7)
1
3 3,
3 1
tSt t
k j j k k
j
net w
(5.8)
( )t t
k kg net (5.9)
The numbers of nodes in each of the input, output and context layers are 6 and 15 in each
hidden layer. The number of neurons in each hidden layers are preferred by iterating
procedures of the training process and found to be 15 as suitable one. The approximation
error (ε) in the output nodes can be minimized using the updated weight factors relation,
i.e., new oldw w w , where ‘ ’, the learning rate is varying from 0 to 1. The
Levenberg-Marquardt algorithm has been applied to the proposed network to estimate the
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
98
crack locations and depth of the structural systems. The sum square error function is
characterized to evaluate the training procedures.
5.5 Application of rule-based modified ERNNs for damage
detection in structure subjected to moving mass
The Elman neural network, a partial recurrent neural network was first recommended by
Elman [230]. The network designed by Elman lies between the conceptions of feed
forward and recurrent network. Like the Jordan networks, the ERNNs have four layers i.e.
input, hidden, output and context layers. The context layer is formed by the feedback links
from the hidden layer. From the context layer, the dynamic memory is provided to the
network. In this analysis, modified ERNNs are proposed for the damage identification in
the structure under moving load. The architecture of the modified ERNNs model is
presented in Figure 5.7. The proposed architecture consists of one input, one output, three
hidden and two context layers. The numbers of neurons in the input and output layers are
6, and those of in each hidden and context layer are 15. The context layer-1 gets
information from the initially hidden layer through feedback links and supplies the outputs
to the primary hidden layer. The context layer-2 receives feedback signals from the
context layer-1 and the outputs of the nodes of the context layer-2 are fed forward to the
first hidden layer. Thus, the context layers-1 and 2, provide dynamic memories to the
network using feedback connections. As the networks have multi-hidden layers, the
feedback connections are provided from the nodes in a hidden layer to the nodes in the
corresponding previous hidden layer. All the nodes have self-recurrent links except those
in the input and output layers. The self-recurrent links in the nodes of the hidden layers
provide more generalizations to the network structure for identification of non-linear
systems.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
99
Where, ‘i=1, 2...N’, ‘N’ is the total number of input nodes. ‘j1= j2= j3=1, 2,..S’, ‘S’ is the
total number of nodes in each of the hidden layer. ‘l1, =l2=1, 2,..T’, ‘T’ is the total number
of nodes in each of the context layer-1and 2.
‘k=1, 2,..O’, ‘O’ is the total number of nodes in the output layer.
‘β’ is the self-recurrent link value in the each node of the context layer-1,context layer-2,
first hidden, second hidden and third hidden layers respectively.
rcd1
rcl2
rcd2
rcl3
rcd3
rcl1
Figure 5.7: Architecture of modified ERNN model
Input layer
First hidden
layer
Output layer
Context layer-1
RD-1
v (m/s)
M (kg)
Z-1
Third hidden
layer
Second hidden
layer
Z-1
Z-1
Context layer-2
Z-1
(6 neurons)
(6 neurons)
(15 neurons) (15 neurons) (15 neurons)
(15 neurons) (15 neurons)
RD-2
RD-3
RD-4
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
100
X1-6 and V1-6 are the values of context nodes in the context layer-1 and 2 respectively.
1
1
t
l and 1
t
l are the net output of the nodes of the context layer-1 at time index ‘t-1’ and
‘t’ respectively.
1
2
t
l and 2
t
l are net the output of the nodes of the context layer-2 at time index ‘t-1’ and
‘t’ respectively.
1
1
t
j and 1
t
j are the net output values of the first hidden layer at time index ‘t-1’ and ‘t’
respectively.
1
2
t
j and 2
t
j are the net output values of the second hidden layer at time index ‘t-1’ and
‘t’ respectively.
1
3
t
j and 3
t
j are the net output values of the third hidden layer at time index ‘t-1’ and ‘t’
respectively.
1t
k
and t
k are the net output values of the output nodes at time index ‘t-1’ and ‘t’
respectively.
The other symbols used in the network have the usual meaning as those in JRNNs model.
From the analysis of the ERNNs model (Figure 5.7), it has been obtained that-
1 1
1 1 1
t t t
l j l (5.10)
1 1
2 1 2
t t t
l l l (5.11)
The net input to the first hidden layer is given by using the following relation-
1 1
1 , 1 1 2 1 2
1
Nt t t t t
j i i j j j l l
i
W w
(5.12)
The net input to the second hidden layer, 1 1
2 1 1, 2 2 3
1 1
St t t t
j j j j j j
j
w
(5.13)
The net input to the third hidden layer or to the network model is given by-
1
3 2 2, 3 3
2 1
St t t
j j j j j
j
w
(5.14)
The netjt = 3
t
j =t
j =f (netjt) (5.15)
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
101
The netkt = 3 3,
3 1
St
j j k
j
w
(5.16)
The net output of the proposed network is given by ( )t t
k kg net (5.17)
The numbers of nodes in each of the input, output layers are 6. 15 neurons are there in
each of the hidden and context layers. The ‘tan-sigmoid’ activation function is employed
in the hidden and context layer, while that in the output layer is ‘purelin’. The proposed
ERNNs have been trained using the Levenberg-Marquardt algorithm to find out the
locations and depth of the cracks on the structure. The ERNNs employs the approximation
error function (ε) in the output nodes to minimize the error value using the updated weight
factors relation, i.e., new oldw w w , where ‘ ’, the learning rate is varying from 0 to 1.
The sum square error function is implemented to evaluate the training process. Like the
JRNNs training process, the same training procedures are carried for the ERNNs model.
5.6 Application of rule-based modified hybridized JRNNs
and ERNNs for multiple damage detection in structure
subjected to moving mass
A novel recurrent neural network structure with the hybridization of both the modified
JRNNs and ERNNs has been presented in Figure 5.8. The hybrid structure is designed by
considering the architectural issue only. The designed structure consists of one input, one
output, three hidden and three context layers. The context layer-1, context layer-2, and
context layer-3 are formed due to the feedback connection from the first hidden layer,
context layer-1 and output layer respectively. Except the nodes in the input and output
layers, all the nodes in the network have self-recurrent connections. The feedback
connections in the hidden layers are provided from the nodes in a hidden layer to the
nodes in the previous hidden layer. Due to the self-recurrent and feedback links, the
dynamic memories are provided to the network for generalization of non-linear systems.
The self-recurrent and feedback links have one time-delay unit each. The hidden layer-1
receives signals from input and context layers-1, 2, 3, and feed forwards it to the hidden
layer-2.
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
102
Figure 5.8: Hybridized architecture of modified JRNN and ERNN
models
Input layer
First hidden
layer
Output layer
Context layer-1
RD-1
v (m/s)
M (kg)
Third hidden
layer
Second hidden
layer
rcd1
rcl2
rcd2
rcl3
rcd3
rcl1
Z-1
Z-1
Context layer-2
Z-1
Z-1
Context layer-3
Z-1
(6 neurons)
(15 neurons) (15 neurons)
(15 neurons) (15 neurons) (15 neurons)
(6 neurons)
(6 neurons)
RD-2
RD-3
RD-4
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
103
The nodes in the output layer produce the desired outputs by receiving information
through feed- forward links from the last hidden layer and due to the self-recurrent links.
1
3
t
l and 3
t
l are net the output of the nodes of the context layer-3 at time index ‘t-1’ and
‘t’ respectively.
The other symbols used in the network have the usual meaning as those in JRNNs and
ERNNs model. From the hybridization of the JRNNs and ERNNs model (Figure 5.8),
1 1
1 1 1
t t t
l j l (5.18)
1 1
2 1 2
t t t
l l l (5.19)
1 1
3 3
t t t
l k l (5.20)
The net input to the first hidden layer is given by using the following relation-
1 1
1 , 1 1 2 1 2 3
1
Nt t t t t t
j i i j j j l l l
i
W w
(5.21)
The net input to the second hidden layer, 1 1
2 1 1, 2 2 3
1 1
St t t t
j j j j j j
j
w
(5.22)
The net input to the third hidden layer or to the network model is given by-
1
3 2 2, 3 3
2 1
St t t
j j j j j
j
w
(5.23)
The netjt = 3
t
j = t
j =f (netjt) (5.24)
The netkt = 1
3 3,
3 1
St t
j j k k
j
w
(5.25)
The net output of the proposed network is given by- ( )t t
k kg net (5.26)
The numbers of nodes in each of the input, output and context layer-3 are 6. 15 neurons
are present in each of the hidden and context-1 and 2 layers. The activation function in the
hidden and context layers is ‘tan-sigmoid’, while that in the output layer is ‘purelin’. The
proposed hybrid structure of the modified JRNNs and ERNNs has been trained using the
Levenberg-Marquardt algorithm to predict the locations and depth of the cracks on the
structure. The approximation error (ε) function has been applied in the nodes of the output
to reduce the error values using the updated weight factors relation, i.e., new oldw w w ,
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
104
where ‘ ’, the learning rate, is varying from 0 to 1. The training processes have been
evaluated using the sum square error function. The training procedures are carried out in
the same manner to those of the JRNNs and ERNNs.
Table 5.1: Test patterns to the RNN model for cracked cantilever beam
Input data to the RNN model Output from the RNN model
RD-1 RD-2 RD-3 RD-4 v (m/s) M (kg) rcl-1 rcl-2 rcl-3 rcd-1 rcd-2 rcd-3
1 1.11 1.21 1.25 6.5 1.5 0.42 0.62 0.72 0.42 0.52 0.32
1 1.06 1.07 1.09 8.5 2.5 0.42 0.62 0.72 0.24 0.25 0.23
1.11 1.16 1.24 1.32 5.9 1.8 0.23 0.57 0.87 0.37 0.43 0.53
1.08 1.13 1.19 1.27 6.3 2.3 0.23 0.57 0.87 0.3 0.4 0.5
0.998 1.23 1.2 1.17 7.3 1.7 0.28 0.48 0.67 0.2 0.3 0.28
1 1.27 1.28 1.21 9.5 2.7 0.28 0.48 0.67 0.22 0.33 0.44
0.996 1.15 1.21 1.19 10 3 0.32 0.53 0.76 0.44 0.33 0.22
1 1.13 1.19 1.17 12 2.8 0.32 0.53 0.76 0.4 0.3 0.2
1 1.12 1.22 1.24 9 2.4 0.44 0.58 0.78 0.44 0.55 0.35
0.997 1.11 1.19 1.21 8.7 2.7 0.44 0.58 0.78 0.4 0.5 0.3
0 100 200 300 400 500 600 700 800 900 10004
4.5
5
5.5
6
6.5
7
7.5
Number of iterations
Su
m s
qu
ared
err
ors
JRNNs
ERNNs
Hybridisation of JRNNs and ERNNs
Figure 5.9: Plot of graph of iterations vs. sum square errors for RNNs methods
model
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
105
Tab
le 5
.2(a
): C
om
par
ison
of
resu
lts
bet
wee
n e
xper
imen
ts a
nd
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tion
of
rela
tiv
e cr
ack l
oca
tion
s (c
anti
lev
er b
eam
)
Exp
erim
enta
l
JRN
Ns
ER
NN
s
Hyb
rid
str
uct
ure
of
JRN
Ns
and
ER
NN
s
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
0.2
49
0.4
502
0
.64
81
0.2
319
0.4
207
0.6
091
0.2
353
0
.42
42
0.6
141
0.2
387
0.4
32
0.6
206
0.4
991
0.6
497
0
.84
81
0.4
643
0.6
087
0.7
908
0.4
714
0
.61
41
0.8
068
0.4
776
0.6
205
0.8
108
0.1
496
0.3
978
0
.62
95
0.1
396
0.3
727
0.5
875
0.1
407
0
.37
6
0.5
908
0.1
428
0.3
801
0.6
006
0.1
981
0.4
685
0
.69
85
0.1
866
0.4
391
0.6
534
0.1
887
0
.44
29
0.6
627
0.1
891
0.4
477
0.6
661
0.2
987
0.5
206
0
.60
03
0.2
808
0.4
853
0.5
596
0.2
809
0
.49
15
0.5
667
0.2
851
0.4
967
0.5
735
0.3
493
0.5
481
0
.71
95
0.3
234
0.5
117
0.6
701
0.3
285
0
.52
01
0.6
82
0
.33
34
0.5
235
0.6
879
0.4
491
0.5
991
0
.74
87
0.4
203
0.5
607
0.6
999
0.4
243
0
.56
61
0.7
047
0.4
281
0.5
714
0.7
145
0.3
982
0.4
994
0
.59
85
0.3
746
0.4
652
0.5
608
0.3
766
0
.47
29
0.5
631
0.3
805
0.4
777
0.5
713
0.2
201
0.6
192
0
.79
88
0.2
047
0.5
784
0.7
487
0.2
074
0
.58
81
0.7
575
0.2
101
0.5
913
0.7
627
0.3
20
0.5
583
0
.85
83
0.2
98
0
.52
44
0.8
072
0.3
012
0
.53
07
0.8
121
0.3
053
0.5
328
0.8
189
Aver
age
per
centa
ge
of
erro
rs
6.5
9
6.4
7
6.4
5
5.6
2
5.3
6
5.4
5
4.4
4
4.4
5
4.5
1
To
tal
per
centa
ge
of
erro
r 6
.5
5.4
7
4.4
3
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
106
Tab
le 5
.2 (
b):
Co
mp
aris
on
of
resu
lts
bet
wee
n e
xper
imen
ts a
nd
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tion
of
rela
tiv
e cr
ack d
epth
(ca
nti
lev
er b
eam
)
Exp
erim
enta
l
JRN
Ns
ER
NN
s
Hyb
rid
str
uct
ure
of
JRN
Ns
and
ER
NN
s
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
0.5
998
0.2
499
0.4
501
0.5
532
0.2
32
0.4
187
0.5
652
0
.23
56
0.4
233
0
.57
51
0.2
378
0.4
314
0.3
006
0.5
498
0.3
991
0.2
786
0.5
123
0.3
718
0.2
841
0
.52
5
0.3
767
0
.28
69
0.5
266
0.3
82
0.1
497
0.4
005
0.5
705
0.1
409
0.3
741
0.5
303
0.1
412
0.3
796
0.5
378
0
.14
31
0.3
817
0.5
466
0.2
001
0.3
501
0.5
396
0.1
879
0.3
269
0.5
05
0.1
887
0
.32
8
0.5
093
0
.19
11
0.3
342
0.5
167
0.2
496
0.2
996
0.4
49
0.2
327
0.2
796
0.4
243
0.2
347
0
.28
15
0.4
254
0
.23
84
0.2
856
0.4
295
0.2
301
0.4
302
0.6
404
0.2
116
0.4
03
0.5
998
0.2
161
0
.40
73
0.6
048
0
.22
01
0.4
112
0.6
135
0.3
193
0.2
698
0.5
297
0.2
984
0.2
52
0.4
965
0.3
007
0
.25
6
0.4
996
0
.30
61
0.2
577
0.5
072
0.3
5
0.4
607
0.5
998
0.3
275
0.4
27
0.5
598
0.3
298
0
.43
45
0.5
657
0
.33
4
0.4
405
0.5
752
0.3
701
0.6
003
0.2
002
0.3
464
0.5
581
0.1
876
0.3
506
0
.56
54
0.1
892
0.3
526
0.5
753
0.1
902
0.5
50
0.5
80
0.5
50
0.5
139
0.5
416
0.5
148
0.5
19
0
.54
64
0.5
216
0
.52
75
0.5
563
0.5
257
Aver
age
per
centa
ge
of
erro
rs
6.5
5
6.5
3
6.4
7
5.4
7
5.5
2
5.5
7
4.4
4
.42
4.3
3
To
tal
per
centa
ge
of
erro
r 6
.51
5.5
2
4.3
8
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
107
Tab
le 5
.3 (
a):
Co
mp
aris
on
of
resu
lts
bet
wee
n e
xp
erim
ents
and
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tio
n o
f re
lati
ve
crac
k l
oca
tio
ns
(sim
ply
su
pp
ort
ed b
eam
)
Exp
erim
enta
l
JRN
Ns
ER
NN
s
Hyb
rid
str
uct
ure
of
JRN
Ns
and
ER
NN
s
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
0.1
071
0.2
142
0.3
214
0.0
996
0.1
995
0.2
99
0.1
008
0
.18
84
0.3
025
0.1
019
0
.20
42
0.3
061
0.1
428
0.2
857
0.3
571
0.1
331
0.2
669
0.3
327
0.1
346
0
.26
92
0.3
367
0.1
361
0
.27
31
0.3
404
0.2
142
0.4
285
0.6
428
0.2
01
0.4
02
0.6
017
0.2
026
0
.40
6
0.6
075
0.2
048
0
.41
04
0.6
138
0.2
142
0.5
714
0.8
571
0.1
997
0.5
352
0.8
039
0.2
029
0
.54
0.8
118
0.2
053
0
.54
76
0.8
199
0.2
5
0.6
071
0.8
928
0.2
335
0.5
675
0.8
381
0.2
361
0
.57
43
0.8
47
0.2
398
0
.58
24
0.8
568
0.1
214
0.4
071
0.5
50
0.1
135
0.3
798
0.5
144
0.1
145
0
.38
42
0.5
206
0.1
159
0
.38
85
0.5
264
0.2
001
0.4
428
0.6
857
0.1
868
0.4
129
0.6
418
0.1
891
0
.41
74
0.6
484
0.1
906
0
.42
23
0.6
559
0.2
357
0.4
142
0.8
142
0.2
205
0.3
859
0.7
646
0.2
23
0
.39
15
0.7
722
0.2
243
0
.39
51
0.7
807
0.2
857
0.5
002
0.7
143
0.2
677
0.4
655
0.6
701
0.2
709
0
.47
24
0.6
752
0.2
733
0
.47
81
0.6
845
0.1
786
0.3
571
0.5
714
0.1
67
0.3
334
0.5
339
0.1
691
0
.33
67
0.5
392
0.1
705
0
.34
18
0.5
465
Aver
age
per
centa
ge
of
erro
rs
6.5
4
6.6
2
6.4
1
5.4
6
5.5
4
5.4
4
4.4
8
4.3
8
4.5
4
To
tal
per
centa
ge
of
erro
r 6
.52
5.4
8
4.4
6
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
108
Tab
le 5
.3(b
): C
om
par
iso
n o
f re
sult
s b
etw
een
ex
per
imen
ts a
nd
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tio
n o
f re
lati
ve
crac
k d
epth
(sim
ply
su
pp
ort
ed b
eam
)
Exp
eri
menta
l
JRN
Ns
ER
NN
s
Hyb
rid
str
uctu
re o
f JR
NN
s an
d
ER
NN
s
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
0.1
495
0.2
997
0.4
495
0.1
395
0.2
798
0.4
19
0.1
41
3
0.2
821
0.4
264
0.1
426
0
.28
5
0.4
284
0.2
007
0.3
995
0.5
003
0.1
863
0.3
756
0.4
667
0.1
886
0
.37
69
0.4
735
0.1
906
0
.38
1
0.4
768
0.5
495
0.5
50
0.5
498
0.5
162
0.5
159
0.5
151
0.5
209
0
.52
13
0.5
22
0.5
27
0
.52
74
0.5
251
0.1
998
0.1
996
0.1
994
0.1
866
0.1
866
0.1
861
0.1
886
0
.18
88
0.1
882
0.1
912
0
.19
06
0.1
901
0.3
499
0.3
489
0.3
502
0.3
273
0.3
274
0.3
265
0.3
309
0
.33
03
0.3
302
0.3
339
0
.33
41
0.3
346
0.4
801
0.3
797
0.4
805
0.4
499
0.3
56
0.4
482
0.4
537
0
.35
83
0.4
544
0.4
591
0
.36
29
0.4
599
0.4
996
0.3
99
0.1
994
0.4
688
0.3
748
0.1
863
0.4
716
0
.37
74
0.1
882
0.4
767
0
.38
13
0.1
904
0.5
303
0.2
505
0.1
785
0.4
962
0.2
336
0.1
679
0.5
015
0
.23
54
0.1
697
0.5
059
0
.23
75
0.1
716
0.3
50
0.4
50
0.5
485
0.3
271
0.4
214
0.5
167
0.3
303
0
.42
67
0.5
216
0.3
343
0
.42
88
0.5
268
0.1
994
0.2
997
0.4
006
0.1
865
0.2
801
0.3
75
0.1
883
0
.28
36
0.3
79
0.1
905
0
.28
52
0.3
826
Avera
ge p
erc
enta
ge o
f err
ors
6
.48
6.3
4
6.4
6
5.5
4
5.5
1
5.3
1
4.5
1
4.5
9
4.4
1
To
tal
perc
enta
ge o
f err
or
6.4
2
5.4
5
4.5
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
109
Tab
le 5
.4(a
): C
om
par
iso
n o
f re
sult
s b
etw
een
ex
per
imen
ts a
nd
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tio
n o
f re
lati
ve
crac
k l
oca
tio
ns
(fix
ed-f
ixed
bea
m)
Exp
eri
menta
l
JR
NN
s
ER
NN
s
Hyb
rid
str
uctu
re o
f JR
NN
s an
d
ER
NN
s
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
0.1
071
0
.22
85
0.3
357
0.0
996
0.2
128
0.3
12
0.1
007
0
.21
48
0.3
154
0.1
018
0
.21
73
0.3
19
0.1
642
0
.30
71
0.4
142
0.1
531
0.2
863
0.3
857
0.1
547
0
.28
93
0.3
901
0.1
562
0
.29
27
0.3
941
0.2
0
.44
28
0.5
857
0.1
862
0.4
141
0.5
462
0.1
887
0
.41
77
0.5
524
0.1
912
0
.42
26
0.5
584
0.2
642
0
.55
71
0.6
857
0.2
468
0.5
217
0.6
408
0.2
495
0
.52
68
0.6
472
0.2
53
0
.53
22
0.6
549
0.3
428
0
.65
71
0.8
357
0.3
208
0.6
16
0.7
845
0.3
243
0
.62
18
0.7
925
0.3
272
0
.62
85
0.7
966
0.2
642
0
.47
85
0.6
928
0.2
474
0.4
475
0.6
495
0.2
502
0
.45
11
0.6
562
0.2
518
0
.45
81
0.6
625
0.3
928
0
.53
57
0.8
214
0.3
682
0.5
02
0.7
694
0.3
725
0
.50
58
0.7
766
0.3
752
0
.51
18
0.7
866
0.4
285
0
.57
14
0.7
142
0.4
022
0.5
36
0.6
681
0.4
066
0
.54
03
0.6
746
0.4
105
0
.54
75
0.6
844
0.5
0
.64
28
0.7
857
0.4
684
0.6
037
0.7
358
0.4
748
0
.60
98
0.7
419
0.4
798
0
.61
67
0.7
501
0.3
214
0
.57
14
0.8
928
0.3
009
0.5
342
0.8
386
0.3
039
0
.54
1
0.8
456
0.3
068
0
.54
58
0.8
548
Avera
ge p
erc
enta
ge o
f err
ors
6
.48
6.4
2
6.4
7
5.4
2
5.5
3
5.5
4
4.4
7
4.4
2
4.5
1
To
tal
perc
enta
ge o
f err
or
6.4
5
5.4
9
4.4
6
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
110
Tab
le 5
.4(b
): C
om
par
iso
n o
f re
sult
s b
etw
een
ex
per
imen
ts a
nd
dif
fere
nt
RN
Ns
met
ho
d f
or
pre
dic
tio
n o
f re
lati
ve
crac
k d
epth
(fi
xed
-fix
ed b
eam
)
Exp
erim
enta
l
JRN
Ns
ER
NN
s
Hyb
rid
str
uct
ure
of
JRN
Ns
and
ER
NN
s
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
rc
d-1
rc
d-2
rc
d-3
0.1
490
0.3
192
0.4
697
0.1
396
0.2
981
0.4
378
0.1
411
0
.30
1
0.4
42
0
.14
22
0.3
041
0
.44
69
0.2
192
0.4
197
0.5
688
0.2
053
0.3
918
0.5
316
0.2
073
0
.39
58
0.5
367
0
.20
92
0.3
995
0
.54
27
0.4
995
0.4
985
0.4
995
0.4
693
0.4
678
0.4
668
0.4
715
0
.47
07
0.4
714
0
.47
58
0.4
767
0
.47
66
0.3
008
0.2
996
0.2
993
0.2
801
0.2
804
0.2
803
0.2
835
0
.28
32
0.2
83
0
.28
57
0.2
863
0
.28
69
0.3
998
0.3
991
0.3
984
0.3
748
0.3
749
0.3
743
0.3
773
0
.37
90
0.3
784
0
.38
16
0.3
824
0
.38
2
0.2
506
0.3
50
0.4
485
0.2
332
0.3
27
0.4
214
0.2
353
0
.33
23
0.4
256
0
.23
91
0.3
351
0
.43
05
0.4
504
0.3
506
0.2
495
0.4
215
0.3
283
0.2
342
0.4
269
0
.33
08
0.2
371
0
.43
12
0.3
354
0
.23
93
0.3
989
0.4
009
0.3
994
0.3
75
0.3
746
0.3
747
0.3
791
0
.37
75
0.3
788
0.3
827
0.3
824
0
.38
32
0.2
50
0.2
988
0.3
505
0.2
336
0.2
808
0.3
271
0.2
365
0
.28
37
0.3
306
0
.23
94
0.2
872
0
.33
5
0.2
607
0.3
807
0.5
60
0.2
426
0.3
551
0.5
257
0.2
457
0
.35
93
0.5
278
0
.24
84
0.3
635
0
.53
69
Aver
age
per
centa
ge
of
erro
rs
6.4
6
6.4
1
6.3
4
5.5
1
5.4
6
5.4
2
4.5
7
4.4
1
4.2
9
To
tal
per
centa
ge
of
erro
r 6
.4
5.4
6
4.4
2
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
111
Tab
le 5
.5(a
): C
om
par
iso
n o
f re
sult
s am
on
g v
ario
us
met
hod
s fo
r pre
dic
tio
n o
f re
lati
ve
crac
k l
oca
tio
ns
(fix
ed-f
ixed
bea
m)
Exp
eri
menta
l
FE
A
Theo
ry
Hyb
rid
str
uctu
re o
f JR
NN
s an
d
ER
NN
s
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
rc
l-1
rc
l-2
rc
l-3
0.3
005
0.5
214
0.8
714
0.2
94
0.5
111
0.8
526
0.2
923
0
.50
69
0.8
489
0
.28
71
0.4
972
0.8
336
0.2
571
0.4
5
0.8
142
0.2
517
0.4
408
0.7
957
0.2
492
0
.43
66
0.7
899
0
.24
58
0.4
303
0.7
785
0.2
004
0.4
428
0.6
928
0.1
956
0.4
331
0.6
785
0.1
941
0
.43
1
0.6
731
0
.19
09
0.4
233
0.6
622
0.1
714
0.4
142
0.6
0
.16
77
0.4
036
0.5
859
0.1
668
0
.40
23
0.5
847
0.1
633
0.3
952
0.5
728
0.2
775
0.5
918
0.8
064
0.2
711
0.5
782
0.7
868
0.2
694
0
.57
7
0.7
84
0
.26
44
0.5
636
0.7
689
0.2
785
0.5
937
0.8
079
0.2
727
0.5
809
0.7
87
0.2
711
0
.57
87
0.7
881
0
.26
58
0.5
65
0.7
728
0.2
796
0.5
945
0.8
083
0.2
729
0.5
803
0.7
89
0.2
72
0
.57
88
0.7
865
0
.26
67
0.5
653
0.7
709
0.2
142
0.5
714
0.8
571
0.2
087
0.5
572
0.8
373
0.2
086
0
.55
66
0.8
337
0
.20
41
0.5
456
0.8
184
0.3
214
0.5
714
0.8
928
0.3
135
0.5
567
0.8
737
0.3
123
0
.55
57
0.8
693
0
.30
73
0.5
448
0.8
529
0.4
285
0.5
714
2
0.7
142
0.4
183
0.5
591
0.6
987
0.4
159
0
.55
55
0.6
941
0
.40
93
0.5
473
0.6
822
Avera
ge p
erc
enta
ge o
f err
ors
2
.24
2.2
7
2.2
8
2.7
7
2.6
9
2.7
4
.52
4.5
8
4.4
7
To
tal
perc
enta
ge o
f err
or
2.2
6
2.7
2
4.5
2
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
112
Tab
le 5
.5(b
): C
om
par
iso
n o
f re
sult
s am
on
g v
ario
us
met
hod
s fo
r pre
dic
tio
n o
f re
lati
ve
crac
k d
epth
(fi
xed
-fix
ed b
eam
)
Exp
eri
menta
l
FE
A
Theo
ry
Hyb
rid
str
uctu
re o
f JR
NN
s an
d
ER
NN
s
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
0.5
51
0.3
508
0
.15
1
0.5
402
0.3
441
0.1
482
0.5
368
0
.34
21
0.1
472
0
.52
6
0.3
357
0.1
444
0.2
01
0.4
21
0
.30
12
0.1
964
0.4
127
0.2
952
0.1
952
0
.40
99
0.2
938
0
.19
16
0.4
006
0.2
869
0.4
43
0.3
302
0
.22
08
0.4
332
0.3
231
0.2
16
0.4
308
0
.32
11
0.2
149
0.4
22
0.3
158
0.2
104
0.4
81
0.3
83
0
.28
11
0.4
697
0.3
742
0.2
748
0.4
683
0
.37
29
0.2
733
0
.45
76
0.3
649
0.2
682
0.2
091
0.3
006
0
.50
51
0.2
039
0.2
943
0.4
946
0.2
039
0
.29
2
0.4
905
0
.19
99
0.2
869
0.4
838
0.1
72
0.3
81
0
.46
08
0.1
684
0.3
72
0.4
499
0.1
672
0
.36
99
0.4
492
0
.16
43
0.3
632
0.4
415
0.4
143
0.2
106
0
.30
17
0.4
049
0.2
054
0.2
944
0.4
025
0
.20
43
0.2
936
0
.39
55
0.2
011
0.2
884
0.1
51
0.2
504
0
.35
04
0.1
474
0.2
444
0.3
417
0.1
471
0
.24
37
0.3
421
0
.14
44
0.2
395
0.3
347
0.3
042
0.4
14
0
.51
11
0.2
968
0.4
04
0.4
999
0.2
962
0
.40
35
0.4
971
0
.29
08
0.3
956
0.4
877
0.6
121
0.6
013
0
.60
1
0.5
989
0.5
877
0.5
892
0.5
964
0
.58
48
0.5
848
0
.58
51
0.5
756
0.5
731
Avera
ge p
erc
enta
ge o
f err
ors
2
.24
2.2
2
2.1
6
2.6
6
2.7
2
.61
4.5
2
4.4
8
4.4
7
To
tal
perc
enta
ge o
f err
or
2.2
2
.65
4.4
9
Chapter 5 Application of Recurrent Neural Networks for Damage Identification
in Structures under Moving Mass
113
5.7 Discussion and Summary
The modified JRNNs, ERNNs and the hybrid structure of both the JRNNs and ERNNs
structures are studied in the present Chapter. The studies have been focused on the
architectural issues. The Levenberg-Marquardt algorithm has been applied to train all the
structures. 600 training patterns are generated for each of the damaged structure to train all
the network models. Some of the testing data for the damaged cantilever beam are
presented in Table 5.1. 1000 numbers of iterations are carried out to train each of the
network models. The relative locations and depth of cracks are determined in different
structures under transit mass with multiple cracks using the recurrent neural networks
methods. The JRNNs, ERNNs, and the hybridisation of both the JRNNs and ERNNs
recurrent neural analysis are applied to predict the locations and depth of cracks using the
Levenberg-Marquardt training algorithm.
The results obtained from the experimental analysis and different RNNs methods are also
presented in Tables 5.2- 4. A graph (Figure 5.9) has been plotted against the number of
iterations and the errors obtained from various RNNs methods to show the accuracies of
the proposed methods. From the analysis of Figure 5.9, it has been observed that the errors
are decreasing with the increase of the number of iterations. The errors in case of the
hybridisation of JRNNs and ERNNs method are less than those in JRNNs and ERNNs
methods. The comparison of results for relative crack locations and depth obtained from
the experiment and different RNNs methods are presented in Tables 5.2, 5.3 and, 5.4 for
cantilever beam, simply supported beam and fixed-fixed beam respectively.
The variation of results between the experiments and JRNNs, ERNNs, and the
hybridisation of both the JRNNs and ERNNs methods are near about with average errors
of 6.46%, 5.47% and 4.4% respectively. The variation of results among the experiments,
FEA and theoretical analyses for the case of fixed-fixed beam is presented in Tables 5.5
(a) and (b). The deviation of results between the experiment and FEA are of near about
2.2%, while those with theoretical analyses are of 2.7% respectively. The Levenberg-
Marquardt training algorithm [263] was applied to train each of the recurrent neural
networks models with 1000 number of iterations. From the analyses of results obtained
from the different recurrent neural network methods, it has been remarked that the
hybridization of the JRNNs and ERNNs yields better results as comparison to JRNNs and
ERNNs methods individually. Thus, the hybridised JRNNs and ERNNs perform better
and proximity results for damage detection in structure under transit mass.
114
Chapter 6
APPLICATION OF STATISTICS
BASED DAMAGE IDENTIFICATION
PROCEDURE FOR STRUCTURES
UNDER TRANSIT MASS
6.1 Introduction
The well-defined fault diagnosis procedures have become a significant problem in modern
mechanical control theory. Early detection of faults in structures provides a potential tool
to execute significant preventing actions. The key features of fault detection methods are
the rule based models which are used for decision making methods. During the normal
operation, the structures under transit mass are suffered from many changing ecological
and operational conditions which can affect the performance of the structural systems. If
the variances produced by the ecological and operational conditions are not appropriately
explained, then the sensitivity of the fault identification approach can be reduced. The
damage detection approaches based on statistical analyses plays a principal role in quality
control and improvement. Statistical analyses are the compilation of methods for preparing
decisions about a method which depend on the investigation of the data contained in a
given sample or population. The statistics based methods present the principal features by
means a product can be sampled, trained and monitored. The information in the sample
data is utilized to improve and control the training and monitoring process. In the present
Chapter, the damage detection procedure is in cast in the perspective of a statistical pattern
recognition problem. The proposed damaged detection method is developed using time
series analysis in the statistical process control (SPC) domain. The analysis has been
carried out using the IBM SPSS 20 software package.
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
115
6.2 Overview of Statistical process control (SPC) method
Statistical process control (SPC), an influential collection of problem solving tools which
is useful to achieve process stability and improve capability through the reduction of
variability. The key features of SPC based variability reduction program is the continuous
improvement on timely programs. In the context of SPC, the problem can be described as
a four-part procedures namely operational evaluation, acquisition of data, feature
extraction and selection of data, and development of statistical model. The present study is
focused on the feature extraction and selection of data, and development of statistical
model. The development of statistical model is concerned with the execution of the
mechanism which can examine the distribution of extracted features to calculate the
damage configuration of the structure. The mechanism implemented in the statistical
model improvement basically categorized in three groups namely classification of group,
regression analysis and outlier identification. The present analysis falls on the regression
analysis category.
The principal role of the SPC technique is to identify the happening of the assignable
defects so that the process can be investigated and necessary preventive action may be
taken before the happening of any nonconforming units. Statistical process control (SPC)
technique can be applied to any procedures. The SPC method has seven important tools
known as ‘the magnificent seven’ namely-
(i) Check sheet. (ii) Steam and leaf plot. (iii) Cause and effect diagram.
(iv) Pareto chart. (v) Scatter diagram. (vi) Defect concentration diagram.
vii) Control chart.
The proposed method is focused on the control chart analysis.
6.3 Construction and analysis of control chart
The control chart analysis is an online monitoring process used for the prediction of
various parameters through proper information. The control chart acts like an important
tool to improve various process parameters. The control chart analysis is the graphical
representation of quality characteristics which has been calculated from a sample number
versus the sample characteristics. The control chart is a tool which can describe the
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
116
statistical process control in an exact and precise manner. The chart is applied for online
inspection of parameter estimation. The representation of a typical control chart is shown
in Figure 6.1.
The data collected from a given sample or population are used to prepare a control chart.
The centre line of the control chart represents the normal average estimated value of the
corresponding control state. If any point that lies outside of the control limit, then that
state is regarded as out of control. So survey and remedial actions are needed to find out
the faults and remove the causes responsible for this kind of behaviour. So control limits
are prepared. According to Montgomery [262], the control limits can be prepared
according to the following equations:
LCL= /2 ww az
s
(6.1)
UCL= /2 ww az
s
(6.2)
Where ‘ s ’ is the sample size, ‘w’ is the sample statistics, ‘µ’ is the mean vale and ‘σw’ is
the standard deviation. The value of ‘ w
s
’ can be tabulated from different sample or
population sizes [262].
The control chart uses the population average ( x ) to supervise the process mean. The
objectives of the control chart are to keep the process in control state. The implementation
of control chart in the SPC method is a significant step for early exclusion of assignable
Sample number
Sam
ple
ch
arac
teri
stic
s
Centre line
Upper control limit (UCL)
Lower control limit (LCL)
Figure 6.1: Architecture of control chart
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
117
causes. There are two types of control charts namely X-chart and R-chart. The current
analogy is dealt with X bar-chart for the control chart analysis.
6.4 Overview of Autoregressive model
It is required to develop a model which can be useful to determine the possibility of a
future value lying between two specific limits. The model may be named as probability or
stochastic model. So for the analyses of the times series, the construction of the model is
required. The present analysis is manifested on the analysis of Auto Regressive (AR)
model. A probability or stochastic model which can be very much useful in the
representation of a specific practically happening series is the Auto Regressive (AR)
model. In the AR model, the present value of the progression is usually represented as a
finite, linear aggregate of past values of the progression and a shock ‘at’. The shocks are
the unsystematic drawing from a linear distribution with zero mean and variance σw2. The
random variables of shocks ‘at, at-1, st-2, at-3.........’ are also known as white noise process.
The representation of a probability model in time series analysis is shown in Figure 6.2.
The probability models based on time series analysis are mostly dependent consecutive
values may be generated from a sequence of independent variables ‘at’ (shocks). The
shocks or white noise are randomly drawn from a fix distribution. These are assumed to be
normal with zero mean and constant variance. The shock or white noise procedure ‘at’ can
be transformed to other process ‘zt’ by a liner filter as in Figure 6.2. The liner filtering
procedures usually gets the summed values of the past shock. So
1 1 2 2 ... ( )t t t tz a a a b (6.3)
Where 2 3
1 2 3( ) 1 ... c
cb b b b b
Shock
or
White noise Outputs
Linear filter
Weights
Figure 6.2: Representation of a probability process as the outputs
from a liner filter
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
118
The representation of a probabilistic or stochastic process as the liner productivity from a
linear filter may be given the equation as:
1 1 2 2 3 3
1
....
or
t t t t c t c t
c
t j t j t
j
z z z z z a
z z a
(6.4)
Where ‘ 1 2 3, , ,.... c ’ are the parameters representing finite set of weights.
Where ‘t, t-1, t-2, ...’ are the uniformly spaced time over 1 2, , ,....t t t t cz z z z and
t tz z .
‘ tz ’ is the deviation of the process from the specified origin. The common normal
procedures as in equation (6.3), permits the process to represent ‘ tz ’ as a weighted sum of
present and past values of the shock or white noise process ‘at’. The present deviation ‘ tz ’
from the mean or level ‘µ’ is regressed on the preceding deviations 1 2, ,....t t t cz z z of the
proposed process.
The method explained by equation (6.3) is named as an AR (c) process with order ‘c’.
This is due to a liner model:
1 1 2 2 3 3 .... c cz x x x x e (6.5)
Here ‘z’ is the dependent parameter to the set of independent parameters, 1 2 3, , ,... cx x x x
and ‘e’ is the error term. The parameter ‘z’ is said to be regressed on the past outputs of it.
So the process is autoregressive. According to equation (6.3), the equation (6.4) can be
also expressed in the equivalent form i.e.
2
1 2(1 ... )c
c t tb b b z a (6.6)
The equation (6.6) can also be written as: ( ) t tb z a (6.7)
1( )t tz b a (6.8)
The AR (c) may be treated as the outcome ‘ tz ’ from the liner filter model having transfer
function ‘ 1( )b ’ with the input ‘ ta ’ to the model.
To avoid over parameterization, it is better to keep the order of the AR model at lower
level.
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
119
6.5 Application of Auto Regressive (AR) model based
method for damage detection in structures subjected to
traversing mass
The damage detection procedures for beam types structures subjected to traversing mass
are carried out using the AR model. For the AR model, 1000 patterns are generated for
each of the structures (cantilever, simply supported and fixed-fixed) at different
configurations of the moving mass structural systems. Out of 1000 patterns, 300 patterns
are for undamaged beam structures while 700 patterns are for damaged structures. The
time- displacement responses histories are determined for all the 1000 numbers of patterns
with 500 numbers of observations intervals. The mean, standard deviation and variances
are computed for each of the pattern. Then the average of the means and standard
deviations are computed by considering all the patterns. The data analysis for the control
chart are done using IBM SPSS software package. Initially, the control charts have been
prepared to know the existence of damages in the structures. The centre line (CL), upper
cutter limit (UCL) and lower control limit (LCL) are computed by considering the
undamaged structures. The detailed analysis of the control charts are represented in
Figures 6(a), 6(b) and 6(c) for the cantilever, simply supported and fixed-fixed beam
structures respectively. If the points lie above the UCL and below the LCL, then it shows
the possibilities of existence of damages in those structures. Selection of group size is an
important issue for preparing X bar control chart.
When the group size is more than one, then the mean of the data within one group is
selected as the chart variable and make one point in the subsequent chart. The selection of
group size has direct impact for the determination of the control limits and thus influences
the sensitivity of the control chart. The statistical process control (SPC) method
particularly based on two types of learning algorithm i.e. unsupervised and supervised.
The SPC method is applied to determine the relative crack locations and crack depth. In
unsupervised learning algorithm for damage detection in structures, the data from the
damaged states are not available, while in supervised learning algorithm, the data are
available.
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
120
Figure 6.3: Data analysis in SPSS windows
-0.5
0
0.5
1
1.5
2
2.5
1 5 9 13 17 21 25
Sam
ple
ob
serv
atio
ns
Sample numbers
sample measure
CL
UCL
LCL
Figure 6.4(a): Control chart for cantilever beam
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
121
In the present analysis, initially supervised learning algorithm is adopted to verify the
accuracy of the proposed method. In the later part, unsupervised learning algorithm is
implemented for fault detection in structure using the statistical process controller. The
autoregressive (AR) method is focused in this analysis.
In the present analogy, the extractions of data are carried out by theoretical-numerical
solution, FEA and experimental methods. The data compression method is used to convert
the time series data from multiple measurement points to single point.
Figure 6.4(b): Control chart for simply supported beam
-0.5
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20 25 30
Sam
ple
ob
serv
atio
ns
Sample numbers
sample observations
CL
UCL
LCL
-0.2
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20 25 30
Sam
ple
ob
serv
atio
ns
Sample numbers
Sample observations
CL
UCL
LCL
Figure 6.4(c): Control chart for fixed-fixed beam
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
122
If ( )i jy t represents the response time histories at ‘n’ measurements points or positions and
sampled at ‘r’ time intervals, then a vector has been formed to define the response
components of subsequent ‘n’ measurement positions. The time history response at a
given time ‘ ( )jt ’ is expressed in the following way-
1 2 3( ) [ ( ) ( ) ( )... ( )]T
i j j j j n jy t y t y t y t y t (6.9)
Where the value of ‘i’ varies from 1 to n. ‘n’ is the number of locations where the
vibration based displacement-time history to be measured along the beam structures. The
displacement-time history data are not only correlated to each other but also to each
other’s previous data.
The equation (6.9) can also be written as:
1 2 3[ , , ... ]T
t t t t nty y y y y (6.10)
Then, the covariance matrix (Ω) of size n n among all the measurement positions over
the entire time intervals is given as: 1
( ) ( )r
T
j j
j
y t y t
(6.11)
Feature extraction is the method which can identify damage sensitive characteristics from
the measured dynamic response of structures that permits one to differentiate between
cracked and uncracked structures. It is difficult to distinguish the time series data from the
damaged and undamaged states by visual inspection. So other features of data extraction
are required for damage identification in structure. In the AR (c) model, the present value
in the time series analysis is represented as the linear combination of the past ‘c’ values.
The autoregressive model with order ‘c’, AR (c), can be expressed as:
1
c
t j t c t
j
y y a
(6.12)
Where ty is the history of time response at time ‘t’, ta is the random error or shock value
at zero mean and constant variance.
The equation (6.12) can also be written as in the form i.e.
1 1 2 2 3 3 ...t t t t c t c ty y y y y a (6.13)
Where, 1 2( , ,... ,... )j c , ‘ j ’ are the coefficient of AR (c) model which represents
the damage sensitive factors.
The values of ‘ j ’ are assessed by fitting the AR (c) model to the displacement- time
response data by applying the Yule-Walker method. [Box et al, 264]. The values of the
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
123
coefficient of the AR model have been calculated with a liner linear least squares
regression using the IMB SPSS software package. The order of the AR (c) model has been
calculated by verifying Gaussianity and randomness of the estimation errors by trial and
error approaches.
The ‘c’ matrices can be formed for each set of the displacement-time histories data which
is the constituent of ‘ j ’ ( n n matrix) and is represented below:
11j 12j 13j 1nj
21j 22j 23j 2nj
n1j n2j n3j nnj
ψ ψ ψ ...ψ
ψ ψ ψ ...ψ
ψ ψ ψ ...ψ
j
(6.14)
The vibration based displacement-time response of the undamaged and damaged
structures are compared indirectly to extract the damage sensitive factors. The damaged
sensitive factors ( j ) have been obtained by fitting the AR (c) model to the displacement-
time history data for each of the response data. The total undamaged data set are divided
into two sample groups namely reference data sample set (SampleR) and healthy data
sample set (Samplehd
), while those of cracked data set are named as Samplecd
. The
SampleR has been utilized to extract damage factors for all data set for future comparison.
The damage sensitive factors for Samplehd
and Samplecd
are also determined. The
extracted damage sensitive factors or coefficient of AR (c) model from the healthy and
damaged states are compared with the reference data sample set (SampleR).
Measuring the variation of damage sensitive factors which occur in the predicted
coefficient of AR model is due to the existence of cracks in the structures. The magnitudes
of variations in the coefficient of AR model or damage sensitive factors are determined
statistically (Fisher Criterion) with respect to the healthy state of the structure. The
coefficient of AR model provides the information about the vibration response data of the
structure. The amount of variation obtained in the coefficients of the AR model at
different locations of structures provides the information about the possible locations of
cracks in the beam structures.
The Fisher Criterion has been applied to calculate the actual variation of coefficient of AR
model or damage sensitive factors for both the damaged and healthy states. The Fisher
criterion is explained as follows:
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
124
2
cd hd
criterion
cd hd
FisherV V
(6.15)
Where ‘ and cd hd ’ are the respective means of damaged sensitive factors of cracked and
healthy states respectively. ‘ and cd hdV V ’ are the respective variances of damaged sensitive
factors of cracked and healthy states respectively. The value of the Fisher criterion will be
more at the possible damage locations. The greater values of the Fisher criterion provide
the information about the sudden increase in the response of the structures and thus able to
identify the damage locations.
Once the identification of the location of cracks is over, then it is desired to quantify the
severities of the identified cracks. The probability density function (PD) and fourth order
statistical moment method (FSMM) are introduced to quantify the severities of cracks
[Wang et al., 246]. The probability density function under Gaussian distribution process
can be articulated as: PD=2( ( )/2 )1
( )2
x xx
x
x e
(6.16)
Where ‘x’ is the amplitude or structural response of the beam structures, ‘ ’ is the
probability density function, ‘ and x x ’ are the mean and standard deviation of the
amplitude ‘x’.
The mathematical expression for the fourth order statistical moment for the amplitude or
structural response ‘x’ can be expressed as:
4 4
4 ( ) ( ) 3
s
x xx x dx (6.17)
The fourth order statistical moment for stiffness vector has been calculated for both the
undamaged and damaged structures. Incorporating the stiffness parameter with an initial
value to equation (4.1), Chapter-4, and the stiffness matrix and the displacement response
of the structures of the predicted crack locations are obtained using the finite element
modelling. The finite element model updating method (Newmark’s integration approach)
has been carried out to measure the cracks severities of the predicted crack element. The
displacement response of the cracked elements obtained from the finite element
modelling, from which the second order central difference approach has been applied to
get the strain values of the corresponding elements of the structures.
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
125
Let 4 ( )k is the simulated statistical moment for displacement response of the structure
with stiffness parameter ‘k’.
‘4
a ’ is the actual statistical moment for the measured displacement response of the
structure.
The residual vectors ( ( )R k ) between the simulated and actual displacement response of
the structure is given by the following equation:
4 4
4
( )( )
( )
akR k
k
(6.18)
The least square method has been applied to minimize the variation between the measured
and actual statistical moments. The information about the stiffness parameters of the
elements are obtained by the least square optimization method. From the element stiffness
parameter, the severities of the damages are quantified.
Table 6.1(a): Comparison of results among FEA, Theoretical and SPC methods for estimation of
relative crack locations (cantilever beam)
FEA Theoretical SPCM
rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3
0.182 0.3256 0.5025 0.1798 0.3214 0.4968 0.1784 0.3195 0.4937
0.2734 0.4214 0.5828 0.2705 0.4169 0.5758 0.2679 0.413 0.5731
0.342 0.485 0.605 0.3384 0.4806 0.5977 0.3346 0.4745 0.5936
0.375 0.506 0.628 0.3715 0.5002 0.6204 0.3665 0.4955 0.6139
0.512 0.624 0.705 0.507 0.6167 0.6961 0.5024 0.6112 0.6911
0.395 0.5525 0.715 0.3914 0.5463 0.7065 0.3857 0.5416 0.7018
0.407 0.576 0.727 0.4038 0.5697 0.7186 0.3989 0.5652 0.7139
0.428 0.526 0.735 0.4223 0.5201 0.7269 0.4195 0.5168 0.7208
0.456 0.639 0.743 0.4511 0.6313 0.7341 0.4462 0.6276 0.7284
0.578 0.676 0.825 0.5703 0.6683 0.8147 0.5649 0.6637 0.8116
Average percentage of error 1.06 1.12 1.19 2.1 1.96 1.85
Total percentage of error 1.12 1.97
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
126
Table 6.1(b): Comparison of results among FEA, Theoretical and SPC methods for estimation of
relative crack depth (cantilever beam)
FEA Theoretical SPCM
rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3
0.157 0.262 0.282 0.1548 0.2591 0.2787 0.154 0.2568 0.2753
0.172 0.243 0.265 0.1694 0.2410 0.2614 0.1683 0.2385 0.2592
0.185 0.268 0.325 0.1831 0.2672 0.3215 0.1815 0.2624 0.3173
0.205 0.285 0.335 0.2031 0.2834 0.3306 0.2007 0.2790 0.3278
0.325 0.408 0.306 0.3222 0.4022 0.3037 0.3178 0.3991 0.2994
0.273 0.312 0.423 0.2704 0.3067 0.4192 0.2672 0.3056 0.4116
0.281 0.35 0.432 0.2785 0.3427 0.4273 0.2763 0.3432 0.4231
0.312 0.423 0.525 0.3076 0.4145 0.5195 0.3062 0.4150 0.5136
0.501 0.521 0.534 0.4953 0.5141 0.5249 0.4936 0.5121 0.5223
0.408 0.415 0.395 0.4046 0.4104 0.3903 0.4028 0.4082 0.3857
Average percentage of error 1.09 1.24 1.15 1.86 1.94 2.27
Total percentage of error 1.16 2.02
Table 6.2(a): Comparison of results among FEA, Theoretical and SPC methods for estimation of
relative crack locations (simply supported beam)
FEA Theoretical SPCM
rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3
0.2178 0.3607 0.4321 0.2154 0.3571 0.4265 0.2142 0.3537 0.4227
0.2285 0.3714 0.4428 0.2259 0.3668 0.4376 0.2245 0.3631 0.4344
0.3042 0.45 0.5142 0.3019 0.4451 0.5068 0.2976 0.4408 0.5033
0.3142 0.3857 0.4571 0.3113 0.3813 0.4518 0.3074 0.3786 0.4486
0.3642 0.4642 0.5164 0.3597 0.4592 0.5106 0.3571 0.4555 0.5073
0.4785 0.5571 0.6021 0.4732 0.5505 0.5947 0.4703 0.5473 0.5914
0.55 0.6928 0.7285 0.5447 0.6846 0.7202 0.5404 0.6794 0.7141
0.5857 0.6857 0.7714 0.5806 0.6778 0.7627 0.5746 0.6727 0.7547
0.6285 0.75 0.8285 0.6211 0.7415 0.8192 0.6161 0.7381 0.8116
0.7014 0.8014 0.8857 0.6917 0.7916 0.8738 0.6866 0.7874 0.8671
Average percentage of error 1.06 1.14 1.21 1.96 1.88 2.02
Total percentage of error 1.13 1.95
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
127
Table 6.2(b): Comparison of results among FEA, Theoretical and SPC methods for estimation of
relative crack depth (simply supported beam)
FEA Theoretical SPCM
rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3
0.173 0.265 0.325 0.1715 0.2624 0.3215 0.1696 0.2594 0.3186
0.328 0.402 0.505 0.323 0.3971 0.5001 0.3223 0.3932 0.4942
0.423 0.429 0.515 0.4179 0.4246 0.5094 0.4151 0.4209 0.5059
0.385 0.455 0.513 0.3815 0.4508 0.5087 0.3774 0.4439 0.5022
0.602 0.607 0.417 0.5969 0.6003 0.4140 0.5896 0.5953 0.4093
0.201 0.306 0.508 0.1987 0.3034 0.5047 0.1965 0.2991 0.4981
0.343 0.521 0.156 0.3395 0.5169 0.1532 0.3355 0.5097 0.1528
0.162 0.276 0.322 0.1601 0.2735 0.3182 0.1585 0.2695 0.3158
0.129 0.368 0.492 0.1274 0.3653 0.4858 0.1263 0.3613 0.4815
0.398 0.565 0.602 0.3932 0.5593 0.5964 0.3907 0.5531 0.5915
Average percentage of error 1.11 0.95 1.05 2.05 2.12 1.87
Total percentage of error 1.03 2.01
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
128
Ex
per
imen
tal
Theo
reti
cal
F
EA
S
PC
M
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
rcl-
1
rcl-
2
rcl-
3
0.2
017
0.3
264
0.3
821
0.1
964
0.3
178
0.3
821
0.1
975
0.3
183
0
.37
38
0.1
938
0.3
146
0.3
668
0.2
281
0.3
013
0.3
642
0.2
214
0.2
928
0.3
642
0.2
246
0.2
946
0
.35
55
0.2
193
0.2
903
0.3
504
0.3
165
0.4
716
0.5
428
0.3
071
0.4
571
0.5
428
0.3
092
0.4
589
0
.52
92
0.3
041
0.4
537
0.5
202
0.3
834
0.4
936
0.5
928
0.3
714
0.4
785
0.5
928
0.3
746
0.4
817
0
.57
88
0.3
678
0.4
749
0.5
683
0.4
183
0.5
363
0.6
214
2
0.4
071
0.5
214
0.6
214
2
0.4
084
0.5
239
0
.60
71
0.4
023
0.5
155
0.5
959
0.4
849
0.6
304
0.6
928
0.4
714
0.6
142
0.6
928
0.4
726
0.6
148
0
.67
61
0.4
673
0.6
046
0.6
634
0.5
516
0.7
219
0.8
428
0.5
357
0.7
014
0.8
428
0.5
381
0.7
043
0
.82
56
0.5
314
0.6
944
0.8
071
0.6
329
0.7
513
0.8
928
0.6
142
0.7
285
0.8
928
0.6
168
0.7
334
0
.87
26
0.6
048
0.7
229
0.8
563
0.6
780
0.7
926
0.8
071
0.6
571
0.7
714
0.8
071
0.6
623
0.7
757
0
.78
84
0.6
489
0.7
633
0.7
739
0.1
910
0.4
979
0.6
714
0.1
857
0.4
857
0.6
714
0.1
864
0.4
858
0
.65
13
0.1
833
0.4
778
0.6
437
Aver
age
per
centa
ge
of
erro
r 2
.87
2.7
8
2.6
7
2.2
8
2.4
2
.37
3.9
6
3.8
1
4.1
To
tal
per
centa
ge
of
erro
r 2
.77
2.3
5
3.9
5
Tab
le 6
.3(a
): C
om
par
ison o
f re
sult
s am
ong E
xper
imen
tal,
FE
A,
Th
eore
tica
l an
d S
PC
met
ho
ds
for
esti
mat
ion
of
rela
tiv
e cr
ack
loca
tio
ns
(fix
ed-f
ixed
bea
m)
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
129
Ex
per
imen
tal
Theo
reti
cal
F
EA
S
PC
M
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
rcd
-1
rcd
-2
rcd
-3
0.2
696
0
.32
38
0.4
296
0
.26
2
0.3
14
0
.41
6
0.2
638
0.3
162
0
.42
08
0.2
589
0.3
112
0.4
133
0.3
528
0
.53
23
0.5
112
0
.34
3
0.5
17
0
.49
6
0.3
439
0.5
193
0
.50
07
0.3
385
0.5
118
0.4
921
0.3
260
0
.43
44
0.5
281
0
.31
7
0.4
22
0
.51
3
0.3
186
0.4
235
0
.51
75
0.3
127
0.4
179
0.5
081
0.5
394
0
.44
18
0.1
213
0
.52
4
0.4
29
0
.11
8
0.5
264
0.4
313
0
.11
88
0.5
173
0.4
231
0.1
166
0.2
207
0
.17
68
0.3
109
0
.21
4
0.1
72
0
.30
2
0.2
152
0.1
728
0
.30
41
0.2
119
0.1
697
0.2
983
0.5
217
0
.62
274
0
.38
55
0.5
07
0.6
05
0
.37
3
0.5
084
0.6
084
0
.37
64
0.5
017
0.5
995
0.3
698
0.2
313
0
.49
55
0.2
552
0
.22
5
0.4
82
0
.24
8
0.2
263
0.4
856
0
.24
95
0.2
218
0.4
761
0.2
451
0.3
286
0
.43
9 0
.52
94
0.3
19
0.4
26
0
.51
5
0.3
207
0.4
301
0
.51
73
0.3
152
0.4
211
0.5
097
0.5
279
0
.24
25
0.4
307
0
.51
3
0.2
35
0
.41
7
0.5
142
0.2
366
0
.42
16
0.5
061
0.2
337
0.4
145
0.4
938
0
.56
26
0.3
787
0
.48
0.5
46
0
.36
8
0.4
816
0.5
491
0
.37
08
0.4
743
0.5
398
0.3
64
Aver
age
per
centa
ge
of
erro
r 2
.83
2.8
9
2.9
5
2.4
2
2.3
1
2.1
4
4.0
2
3.9
1
3.8
6
To
tal
per
centa
ge
of
erro
r 2
.9
2.2
9
3.9
3
Tab
le 6
.3(b
): C
om
par
ison o
f re
sult
s am
ong E
xper
imen
tal,
FE
A, T
heo
reti
cal
and S
PC
met
ho
ds
for
esti
mat
ion
of
rela
tiv
e cr
ack
dep
th (
fix
ed-f
ixed
bea
m)
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
130
6.6 Discussion and Summary
The present Chapter elaborates a noble damage detection procedure for structures under
traversing mass using statistical process control approach. The method has been divided
into two parts namely training and monitoring. The concept of control chart is introduced
to know the existence of damages in the structures. The control charts for the cantilever,
simply supported and fixed-fixed beam structures are represented in Figures 6.4(a), 6.4(b),
and 6.4(c) respectively. From the analysis of control chart, one can confirm whether the
structure is in damaged or healthy states.
The statistical process control method based on the perception of autoregressive (AR)
model is analyzed in time series domain for the fault detection in damaged structures. The
coefficients of AR model or damage sensitive features are determined using IBM SPSS
software package which provides the information about the vibration response data of the
structure. The magnitude in the actual variation in the AR model coefficients have been
computed statistically (Fisher Criterion) for both the damaged and healthy states of the
structures. The greater Fisher criterion values provide the information about the abrupt rise
in the response of the structures and thus able to recognize the relative crack locations.
After the damage localization process is over, and then the quantification of crack
severities is started. The stiffness vectors for each element of the damaged and undamaged
structures are determined using the fourth order statistical moment method. The central
differential scheme has been approached to calculate the strain values of the predicted
crack elements. The residuals vector have been obtained by considering the simulated
statistical moment with stiffness parameter and the actual statistical moment for the
measured displacement response of the structures. The information about the stiffness
parameters are obtained by minimizing the two statistical moments using the least square
algorithm. From the stiffness parameters, the damage severities are calculated. The results
obtained regarding the relative crack locations and crack depth from the statistical process
control approaches are represented in Tables 6.1 to 6.3 with FEA and theoretical methods
to verify the exactness of the proposed method. The comparison of results among various
methods for the determination of relative crack location are represented in Table 6.1(a) for
cantilever beam, Table 6.2(a) for simply supported beam and Table 6.3(a) for fixed-fixed
beam respectively, while those for relative crack depth the results are represented in
Tables 6.1(b), 6.2(b), 6.3(b) for cantilever, simply supported and fixed-fixed beam
Chapter 6 Application of Statistics Based Damage Identification Procedure for
Structures under Transit Mass
131
respectively. It has been observed that the results obtained from SPC approach vary with
an average variation about 2.02 % and 3.93% with FEA and experiments respectively. So
the proposed method can be effectively applied to identify and quantify the cracks in the
structures.
132
Chapter 7
COMBINED HYBRID NEURO-
AUTOREGRESSIVE MODEL FOR
FAULT DETECTION IN BEAM
STRUCTURES SUBJECTED TO
MOVING MASS
7.1 Introduction
The use of mathematical model to explain the characteristics of physical phenomenon can
be well established. It is sometimes feasible to develop a method based on knowledge
based physical laws. The knowledge based may be some pattern recognition problems. In
fact no phenomenon is completely deterministic it’s because of the occurrence of
unknown factors. The problems based on statistical pattern recognition have recently
appeared as a promising tool to for automatic structural damage estimation. It is required
to implement some knowledge based theory in statistics based method and neural
networks approach for condition monitoring of structures. Numerous patterns recognition
algorithms based on statistical process control and neural network approaches are
developed for the automatic damage identification in structures. In this study, a combined
neural networks (Hybridisation of JRNNs and ERNNs) and autoregressive process
(Statistics process control method) based approach is developed for the structural damage
identification for beam type structures under moving mass. The efficiency and exactness
of the developed approach are also analyzed.
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
133
7.2 Development of combined hybrid neuro-autoregressive
model for damage detection in beam type structures
subjected to moving mass
The integrated approach, hybrid neuro autoregressive model, is proposed for structural
damage detection based on the analyses of the autoregressive model and recurrent neural
networks. The autoregressive process is based on the domain of statistical process control
(SPC) method. Before the formulation of the combined hybrid-neuro autoregressive model
based approach, the RNNs (Chapter-5) and autoregressive (Chaptr-6) based methods are
developed. The objective of combining the two models (neural networks and
autoregressive models) is to refine the results obtained from the individual model and to
enhance the accuracy of the new developed method. The simple architecture of the
combined hybrid neuro-autoregressive model is represented in Figure 7.1.
Figure 7.1: Architecture of Hybrid neuro autoregressive model
AR Model
RCL-1
RCD-1
RCL-2
RCD-2
RCL-3
RCD-3
Displacement time histories
data of structures
NNs model
RCL-1
RCD-1
RCL-2
RCD-2
RCL-3
RCD-3
RD-1
RD-2
RD-3
RD-4
v (m/s)
M (kg)
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
134
In the proposed model, the autoregressive and NNs models are combined together to carry
out the fault detection procedures. Initially, the autoregressive analysis is carried out to
determine the relative crack locations and depth. The procedures to determine the relative
crack locations and depth using the autoregressive model are already explained in
Chapter-6. The same procedures are maintained here to find out the crack locations and
depth. Once the outputs are obtained from the autoregressive model, then these are
immediately fed to the neural network model as input parameters. The present neural
networks model has now some additional inputs apart from their original inputs.
The neural networks model is a recurrent neural network model. The architecture of the
present network model is same as that of the previous network model (Figure 5.8,
Chapter-5); the only difference is the additional inputs from the autoregressive model. The
present RNNs model, the hybridised model of JRNNs and ERNNs has 12 numbers of
input parameters to the network model. The other architectural issues and mechanism are
same as that in Figure 5.8 (Chapter-5).
The net input to the proposed network model is-
1
3 2 2, 3 3
2 1
St t t
j j j j j
j
w
(Chapter-5, equation- 5.23).
The netjt = 3
t
j = t
j =f (netjt) (Chapter-5, equation-5.24)
The netkt = 1
3 3,
3 1
St t
j j k k
j
w
(Chapter-5, equation-(5.25)
The net output of the proposed network is given by- ( )t t
k kg net . (Chapter-5, equation-
5.26)
Where, the symbols have their usual meanings as in Chapter-5.
The proposed neural network model has been also trained with Levenberg-Merquardt
algorithm with the same training procedures as in Chapter-5. The training processes have
been carried out using the sum square error function. The various relative crack positions
and depth are calculated.
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
135
Table 7.1(a): Comparison of results among FEA, Theoretical and hybrid approach of RNNs, and
SPC methods for prediction of relative crack locations (cantilever beam)
FEA Theoretical Hybrid RNNs and SPC
rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3
0.1625 0.3163 0.4318 0.1607 0.3124 0.4264 0.1600 0.3108 0.4241
0.2524 0.3928 0.4717 0.2500 0.3881 0.4664 0.2482 0.3855 0.4632
0.3672 0.5254 0.6436 0.3630 0.5198 0.6357 0.3619 0.5179 0.6335
0.4017 0.5015 0.6525 0.3967 0.4957 0.6447 0.3962 0.4938 0.6416
0.2225 0.4525 0.6727 0.2200 0.4474 0.6643 0.2189 0.4454 0.6622
0.4216 0.5272 0.8025 0.4176 0.5216 0.7938 0.4151 0.5182 0.7911
0.4415 0.582 0.743 0.4367 0.5759 0.7345 0.4354 0.5733 0.7307
0.517 0.584 0.646 0.5111 0.5776 0.6384 0.5088 0.5746 0.6344
0.6016 0.7215 0.8125 0.5962 0.7133 0.8027 0.5931 0.7092 0.7999
0.3726 0.5775 0.7755 0.3691 0.5709 0.7655 0.3666 0.5677 0.7619
Average percentage of error 1.06 1.12 1.23 1.52 1.63 1.65
Total percentage of error 1.14 1.6
Table 7.1(b): Comparison of results among FEA, Theoretical and hybrid approach of RNNs, and
SPC methods for prediction of relative crack depth (cantilever beam)
FEA Theoretical Hybrid RNNs and SPC
rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3
0.1425 0.2515 0.2728 0.1406 0.2488 0.2697 0.1398 0.2474 0.2681
0.2015 0.283 0.302 0.1991 0.2803 0.2988 0.1980 0.2786 0.2969
0.2065 0.305 0.3565 0.2041 0.3022 0.3522 0.2032 0.3007 0.3509
0.2455 0.2915 0.152 0.2431 0.2882 0.1501 0.2419 0.2875 0.1497
0.308 0.408 0.505 0.3047 0.4038 0.4989 0.3037 0.4022 0.4964
0.416 0.453 0.486 0.4112 0.4474 0.4814 0.4096 0.4465 0.4782
0.512 0.555 0.578 0.5059 0.5485 0.5712 0.5039 0.5463 0.5696
0.4565 0.528 0.544 0.4513 0.5217 0.5379 0.4496 0.5200 0.5354
0.3915 0.448 0.576 0.3867 0.4428 0.5703 0.3859 0.4414 0.5677
0.335 0.412 0.562 0.3307 0.4076 0.5559 0.3300 0.4063 0.5528
Average percentage of error 1.17 1.08 1.11 1.56 1.47 1.6
Total percentage of error 1.12 1.54
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
136
Table 7.2(a): Comparison of results among Experimental, FEA, Theoretical and hybrid approach
of RNNs, and SPC methods for prediction of relative crack locations (simply supported beam)
Experimental Theoretical FEA Hybrid RNNs and
SPC
rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3
0.2250 0.3750 0.4464 0.2185 0.3643 0.4350 0.2199 0.3669 0.4360 0.2166 0.3605 0.4315
0.2429 0.3857 0.4500 0.2362 0.3758 0.4376 0.2371 0.3778 0.4397 0.2336 0.3711 0.4344
0.3071 0.4929 0.5571 0.2984 0.4796 0.5410 0.2997 0.4818 0.5439 0.2971 0.4763 0.5384
0.3929 0.4857 0.5286 0.3806 0.4734 0.5141 0.3839 0.4742 0.5158 0.3789 0.4688 0.5112
0.4357 0.5286 0.6071 0.4242 0.5133 0.5911 0.4254 0.5163 0.5938 0.4194 0.5098 0.5879
0.5071 0.5857 0.6571 0.4926 0.5680 0.6401 0.4960 0.5715 0.6429 0.4876 0.5630 0.6345
0.5929 0.6643 0.7714 0.5768 0.6450 0.7508 0.5793 0.6512 0.7557 0.5736 0.6392 0.7439
0.6857 0.7571 0.8286 0.6666 0.7363 0.8059 0.6713 0.7416 0.8124 0.6622 0.7292 0.7976
0.3143 0.6286 0.8071 0.3053 0.6111 0.7841 0.3070 0.6164 0.7919 0.3036 0.6046 0.7779
0.2357 0.5214 0.8000 0.2290 0.5074 0.7790 0.2306 0.5098 0.7829 0.2276 0.5021 0.7706
Average percentage of
error
2.83 2.73 2.68 2.27 2.18 2.17 3.53 3.67 3.44
Total percentage of error 2.74 2.2 3.54
Table 7.2(b): Comparison of results among Experimental, FEA, Theoretical and hybrid approach
of RNNs, and SPC methods for prediction of relative crack depth (simply supported beam)
Experimental Theoretical FEA Hybrid RNNs and SPC
rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3
0.252 0.324 0.462 0.2444 0.3151 0.4492 0.2466 0.3169 0.4530 0.2433 0.3132 0.4449
0.334 0.576 0.482 0.3254 0.5608 0.4683 0.3271 0.5648 0.4712 0.3213 0.5550 0.4643
0.375 0.405 0.315 0.3651 0.3944 0.3058 0.3661 0.3970 0.3076 0.3613 0.3910 0.3044
0.156 0.526 0.294 0.1519 0.5127 0.2862 0.1522 0.5147 0.2869 0.1508 0.5085 0.2828
0.452 0.357 0.589 0.4396 0.3481 0.5729 0.4431 0.3493 0.5744 0.4359 0.3449 0.5664
0.551 0.426 0.217 0.5357 0.4150 0.2109 0.5404 0.4164 0.2119 0.5303 0.4114 0.2088
0.268 0.378 0.465 0.2608 0.3676 0.4520 0.2623 0.3690 0.4540 0.2582 0.3648 0.4479
0.295 0.472 0.496 0.2873 0.4594 0.4819 0.2883 0.4606 0.4851 0.2844 0.4553 0.4785
0.445 0.353 0.592 0.4335 0.3431 0.5750 0.4345 0.3447 0.5791 0.4293 0.3403 0.5703
0.512 0.202 0.401 0.4990 0.1965 0.3893 0.5009 0.1976 0.3920 0.4944 0.1950 0.3858
Average percentage of error 2.68 2.74 2.83 2.18 2.2 2.27 3.57 3.46 3.68
Total percentage of error 2.75 2.22 3.57
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
137
Table 7.3(a): Comparison of results among FEA, Theoretical and hybrid approach of RNNs, and
SPC methods for prediction of relative crack locations (fixed-fixed beam)
FEA Theoretical Hybrid RNNs and SPC
rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3 rcl-1 rcl-2 rcl-3
0.1486 0.2179 0.3257 0.1471 0.2158 0.3229 0.1461 0.2142 0.3205
0.1843 0.2543 0.3236 0.1822 0.2518 0.3197 0.1811 0.2508 0.3189
0.2179 0.2971 0.3771 0.2156 0.2945 0.3731 0.2146 0.2930 0.3710
0.2707 0.3229 0.4129 0.2675 0.3196 0.4085 0.2665 0.3186 0.4063
0.4300 0.5025 0.5775 0.4248 0.4969 0.5710 0.4225 0.4956 0.5690
0.4429 0.5500 0.6143 0.4383 0.5443 0.6083 0.4356 0.5429 0.6048
0.4571 0.5929 0.7143 0.4518 0.5864 0.7062 0.4493 0.5837 0.7034
0.5571 0.6357 0.8214 0.5502 0.6285 0.8129 0.5477 0.6257 0.8079
0.3357 0.5571 0.7714 0.3318 0.5477 0.7635 0.3304 0.5485 0.7605
0.3786 0.6571 0.7286 0.3725 0.6497 0.7203 0.3727 0.6474 0.7171
Average percentage of error 1.18 1.1 1.06 1.66 1.45 1.56
Total percentage of error 1.11 1.56
Table 7.3(b): Comparison of results among FEA, Theoretical and hybrid approach of RNNs, and
SPC methods for prediction of relative crack depth (fixed-fixed beam)
FEA Theoretical Hybrid RNNs and SPC
rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3 rcd-1 rcd-2 rcd-3
0.2520 0.3240 0.4620 0.2495 0.3211 0.4580 0.2479 0.3190 0.4541
0.3340 0.5760 0.4820 0.3305 0.5704 0.4775 0.3284 0.5664 0.4755
0.3750 0.4050 0.3150 0.3708 0.4010 0.3111 0.3685 0.3979 0.3103
0.1560 0.5260 0.2940 0.1541 0.5195 0.2905 0.1532 0.5173 0.2894
0.4520 0.3570 0.5890 0.4467 0.3525 0.5820 0.4445 0.3515 0.5809
0.5510 0.4260 0.2170 0.5443 0.4208 0.2145 0.5425 0.4197 0.2141
0.2680 0.3780 0.4650 0.2651 0.3737 0.4604 0.2636 0.3720 0.4584
0.2950 0.4720 0.4960 0.2923 0.4665 0.4910 0.2900 0.4644 0.4888
0.4450 0.3530 0.5920 0.4405 0.3487 0.5859 0.4384 0.3480 0.5829
0.5120 0.2020 0.4080 0.5071 0.1998 0.4036 0.5043 0.1993 0.4024
Average percentage of error 1.07 1.12 1.06 1.63 1.55 1.46
Total percentage of error 1.08 1.54
Chapter 7 Combined Hybrid Neuro-Autoregressive Model for Fault Detection
in Beam Structures Subjected to Moving Mass
138
7.3 Discussion and Summary
The implementation of damage detection scheme is often consigned to as structural health
monitoring process. This implementation process involves the characterization of potential
damage scenarios for the structural system. The extraction and analysis of damage features
determine the actual state of the system. Many damage assessment approaches have been
proposed. In this analysis a combined hybrid neuro-autoregressive model based approach
has been developed to locate the positions and quantify the severities of cracks in the
structure. The purpose of the combining the two methods is to improve the results
obtained from the individual methods.
The architectural scheme of the proposed integrated method is represented in Figure 7.1.
The autoregressive model is trained with the displacement time history data obtained from
the response of the structure. The output parameters from the autoregressive models are
fed to the RNNs model as input parameters along with the network’s normal input
parameters. The training procedure has been performed by employing the Levenberg-
Merquardt with 1000 number of iterations. The results obtained regarding the relative
crack depth from the present integrated approach are represented in Tables 7.1(a) for
cantilever,7.2(a) for simply supported and 7.3(a) for fixed-fixed beam structures
respectively, while those related to relative crack depth are represented in Tables
7.1(b),7.2(b) and 7.3(b) for cantilever, simply supported and fixed-fixed structures
respectively.
From analyse of results, it has been observed that the average variation of results between
the experimental and integrated approach of RNNs, and AR methods are near about 3.5%,
while those between the FEA and integrated approach of RNNs, and AR methods are near
about 1.5%. It has been concluded that the integrated approach of combined hybrid neuro-
autoregressive method predicts better results as comparison to RNNs and SPC
(autoregressive process) methods individually.
139
Chapter 8
EXPERIMENTAL ANALYSIS OF
DAMAGED STRUCTURES
SUBJECTED TO TRANSIT MASS
8.1 Introduction
Recently, structures under the action of moving objects are an emergent topic in the field
of structural dynamics. In most of the cases, the theoretical analysis or FEA does not
ensure the actual responses of structures precisely. To overcome the above limitations, real
time experimental techniques are necessary to study the actual responses of the structures.
The current Chapter addresses the detailed experimental procedures to determine the
responses of the beam structures subjected to moving mass at different end conditions.
8.2 Experimental Procedure
An experimental set-up has been developed in the laboratory for the moving mass-
structure systems. To verify the theoretical-numerical and FEA solutions, experiments
have been conducted for the different beam structures (cantilever, simply supported and
fixed-fixed beams) under traversing mass in the laboratory. The experimental set-up for
the transit mass-structure system are shown in Figures 8.1 (cantilever beam), 8.2 (simply
supported beam) and 8.3 (fixed-fixed beam). The different components of the
experimental set-up with machine specification are illustrated in Table 8.1. The
experimental procedures are carried out to determine the responses of the structures at
different position of the transit mass and also at the specified locations of the beam
specimen. The different components of the experimental set up are arranged as in the
Figures 8.1, 8.2 and 8.3. The mild steel beam specimens are considered for the
experimental analysis. The dimensions of the beam structures remain same as those of
theoretical-numerical analysis. The cracks are made at the appropriate positions of the
beam using wire EDM machine (Figure 8.4).
Chapter 8 Experimental Analysis of Damaged Structures Subjected to Transit Mass
140
Sen
sors
Mo
vin
g m
ass
Dis
pla
y u
nit
Var
iac
AC
mo
tor
Mic
ro c
ontr
oll
er
Bre
ad b
oar
d
Ro
pe
Fig
ure
8.1
: E
xper
imen
tal
set
up f
or
canti
lever
bea
m
Chapter 8 Experimental Analysis of Damaged Structures Subjected to Transit Mass
141
The mass was allowed to slide on the cracked beam structure by connecting it one end of
the rope and another end of the rope was appended to the pulley. The pulley which in turn
was fixed to the shaft of the motor as in the Figures 8.1, 8.2 and 8.3. The main power
supply is A.C. So that power is supplied to the components of the experimental set-up
such as motor, variac, monitor, data acquisition unit and micro controller. The traversing
mass is allowed to slide across the beam structures without slipping. The length between
Figure 8.2: Experimental set up for simply supported beam
Display unit
Variac
AC motor
Rope
Moving mass
Sensors
Micro controller
Bread board
Figure 8.3: Experimental set up for fixed-fixed beam
Chapter 8 Experimental Analysis of Damaged Structures Subjected to Transit Mass
142
the traversing mass and pulley was adjusted in a suitable manner such that there is no
slackness in the rope. The motor has been set with screws in a wooden tool in a proper
way that it will not be dislocated during the operation. The knob of the variac is adjusted
accurately to get the requisite constant speed of the traversing object and the best probable
precision in measurements. Intensive cares have been taken to place the variac knob to
obtain the required speed of the traversing mass. The speed of the traversing object is
assumed to be uniform while moving across the beam. Ultrasonic sensors are placed
below the beam structures at different positions through microcontroller and the
microcontroller is connected to the monitor (Display unit) through data acquisition unit.
The dynamic deflections of the beam structures are recorded through the sensors and
presented on the monitor. The average readings of dynamic deflections of the beam
structures at the different positions of the traversing mass are determined through the
sensors. Several numbers of tests are conducted to find out the deflections of the damaged
beam structures (cantilever, simply supported and fixed-fixed) with numerous speeds and
weight of the traversed mass.
Figure 8.5: Microcontroller (Audrino MEGA)
Figure 8.4: Ultrasonic sensor
Chapter 8 Experimental Analysis of Damaged Structures Subjected to Transit Mass
143
Table 8.1: Specification of different components of experimental set up
S.N Items Specification
1 Power supply AC 220, 50Hz
2 Motor Type AC, 230V, 0.5 A,1/2 HP, RPM 1400, Type- M32.
Shaft diameter= 2.8 cm
3 Variac Input -230V, 50-60Hz,Output – 0-270V
4 Bread board Small Size Bread Board
5 Communication USB connection Serial Port
6 Microcontroller
Arduino MEGA , ATmega2560, 256 KB (ATmega2560)
Operating Voltage 5V, SRAM 8 KB (ATmega328)
Input Voltage (Recommended) 7-12V
Input Voltage (Limits) 6-20V, Digital Input Pins 54 (of Which 15 Provide
PWM Output), Analog Input Pins 14
7 Sensors Ultrasonic Range Finder Sensor Distance Measuring Range: 1cm to 400cm
8 Moving Mass 1 kg and 2 kg.
9 Specimen Mild steel
10 Cantilever beam Size=100cm×3.9cm×0.5cm, 1,2,3 =0.6, 0.25, 0.45 and 0.3, 0.55, 0.4
1,2,3 =0.25, 0.45, 0.65 and 0.5, 0.65, 0.85. v = 4.38 m/s and 5.73 m/s
11 Simply supported
beam
Size=140cm×4.9cm×0.5cm, 1,2,3 =0.2,0.3,0.4 and 0.35,0.45,0.55.
1,2,3 =0.2857, 0.5, 0.7143 and 0.1786, 0.3571, 0.5714. v = 4.38 m/s and
5.73 m/s
12 Fixed-fixed
beam
Size=140cm×4.9cm×0.5cm, 1,2,3 =0.2,0.35,0.45 and 0.3,0.5,0.55.
1,2,3 =0.1429, 0.3214, 0.5357 and 0.25, 0.4286, 0.7143. v = 5.12 m/s and
6.17 m/s
13 Vibration shaker Bruel and Kjaer, Frequency range-5 to 10 KHz, Maximum bare table
acceleration-700m/s2.
14 Delta Tron
Accelerometer Frequency range -1 to 10 KHz. Sensitivity-10mv/g to 500mv/g.
Figure 8.6: Damaged portion of beam
Chapter 8 Experimental Analysis of Damaged Structures Subjected to Transit Mass
144
8.3 Discussions and summary
The experiments are carried out for the different beam structures under moving mass in
the laboratory. The responses of the structures at different damage scenarios are studied.
The dynamic deflection of the structures at different positions of the moving object and
specified locations are recorded through the sensors and Audrino micro controller. The
measured beam deflection from the experimental analysis are compared with those of
theoretical and FEA. In the experimental verifications, similar observations are obtained
regarding the responses of the structures as those in theoretical analysis and FEA. The
variations of results obtained from experimental analysis are with an average error of near
about 3% and 5% with FEA (Chapter-4) and theoretical analysis (Chapter-3) respectively.
So the proposed methods in the theoretical-numerical solutions and FEA converge well
with the experimentations. The relative crack positions and crack depth of the damaged
beam structures from the experimental analysis are also converged well with the different
methods like FEA, theoretical, RNNs, SPC, and integrated approach of the RNNs and
SPC.
Figure 8.8: Bread board
Figure 8.7: Variac
145
Chapter 9
RESULTS AND DISCUSSION
9.1 Introduction
The present Chapter elaborates the descriptions about the systematic procedures followed
in this thesis. The responses of the structures subjected to traversing mass are studied. The
numerical methods like Runge-Kutta and Newmark’s integration methods are applied to
study the dynamic behaviour of the structures under moving mass. The different damage
detection procedures such as JRNNs, ERNNs, combined JRNNs and ERNNs, AR process
and the combined hybrid process of JRNNs, ERNNs, and AR methods are discussed here.
The efficiencies and exactness of each method are explained.
9.2 Analysis and results of different adopted methods
The present study has been started with the review of different approaches to analyze the
response of different structures and ended with different methods to identify faults in
structures. The present dissertation is divided into nine Chapters. In each Chapter, the
different adopted methodologies are explained.
The motivation behind the research work, the objectives and novelty of the proposed
work, the layout of the entire thesis are explained in Chapter-1. The Chapter-2 (literature
review) explains about the motivate works carried out by different researcher, engineers
and scientist in the field of vibration and structural dynamics. The applications of various
numerical, FEA, experimental and soft computing methods have been also studied. From
the analysis of literature reviews, it gives us the knowledge gap between the previous and
present studies. Keeping in mind the knowledge gap between the past and present
analysis, the different techniques such as RNNs, SPC and the integrated approach of
RNNs, and SPC methods are applied for the crack detection in structures. The current
thesis is mainly organised into two categories i.e. determination of responses of structures
and detection of faults in structures.
In Chapter-3, the theoretical-numerical solutions of the multi-cracked structures subjected
to moving mass are analyzed. The definition and formulation of the proposed problem is
Chapter 9 Results and Discussion
146
also discussed here. The representation of multi-cracked cantilever beam (Figure 3.1),
simply supported beam (Figure 3.16), fixed-fixed beam (3.31) under traversing mass are
shown. The dynamic responses of the structures are investigated at different damage
configurations, moving mass and moving speeds of the structural systems. The solution of
the governing equation (3.18) has been solved with a numerical procedure of fourth order
Runge-Kutta method in MATLAB environment. The detailed response analyses of the
structures subjected to transit mass have been explained in Figures 3.2-3.11 (cantilever
beam), 3.17-3.26 (simply supported beam) and 3.32-3.41 (fixed-fixed beam). The 3-D
graphs for cantilever beam (Figures 3.12-3.15), simply supported beam (Figures 3.27-
3.30), fixed-fixed beam (Figures 3.43-3.45) are also explained at varying moving mass
and speed. The existence of cracks can be also known from the measured dynamic
deflections of the structures (Figures 3.46, 3.47 and 3.48). From the response analyses of
the structures, the consequence of parameters like crack locations, crack depth, moving
speed and moving mass are investigated on the response of the structures. The comparison
studies have been also carried out between undamaged and damaged structures under
traversing object.
In Chapter-4, the finite element analysis (FEA) using commercial ANSYS
WORKBENCH 2015 has been applied to determine the responses of the beam type’s
structures under transit mass. The transient dynamic analysis approach using the full
method is adopted in ANSYS WORKBENCH 2015. The computational approach
implemented in FEA is the Newmark’s integration method. The different steps involved in
the full method transient dynamic analysis are also explained. Before analyzing the
transient dynamic analysis, the modal analyses are performed to find out the natural
frequencies and mode shapes of the structures. The transit mass interaction dynamics of
the structural system is shown in Figure 4.5 (cantilever beam) and the magnified view of
the crack zone is shown in Figure 4.3. The frequencies ratios of the damaged structures are
represented in Tables 4.1, 4.4 and 4.7 for the cantilever, simply supported and fixed-fixed
beam structures. The different modal behaviour of the structures is also represented in
Figure 4.4(cantilever beam), Figure 4.7(simply supported beam) and Figure 4.9 (fixed-
fixed beam). Similar observations are made regarding the response of the structures in
FEA as those in numerical analysis.
In Chapter-8, the experimental models have been developed for each of the structure
subjected to traversing object in the laboratory. The laboratory set-ups for the structural
Chapter 9 Results and Discussion
147
systems under moving mass are shown in Figures 8.1, 8.2 and 8.3 for cantilever, simply
supported and fixed-fixed structures respectively. The different equipments of the
experimental set-ups, their specifications and functions are described in Table 8.1. The
experimental procedures have been already explained in Chapter-8. The responses of the
structures are also determined by laboratory tests. The purpose of the laboratory tests is to
verify the accuracy and exactness of FEA and numerical analysis.
The detailed analyses regarding the response of the structures have been illustrated in
Chapters 4, 5 and 8. The accuracy of each method has been compared with each other.
The comparison of results among experiments, FEA and numerical analysis for the
response of structures are represented in graphical way in Figure 4.6 (damaged cantilever
beam), Figure 4.8 (damaged simply supported beam) and Figure 4.10 (damaged fixed-
fixed beam). The variation of results between the experimental and numerical analyses are
expressed in Tables 3.1,3.2 and 3.3 for cantilever, simply supported and fixed-fixed beam
structures under moving mass respectively. Similarly, the disparity of results among the
laboratory tests, FEA and numerical analysis are demonstrated in Tables 4.2-4.3
(cantilever beam), Tables 4.5-4.6 (simply supported beam) and Tables 4.8-4.9 (fixed-fixed
beam. From analyses of results obtained from the different methods, it has been observed
that there are variations of results near about 5% between the experimental and numerical
analyses, while those between the experimental and FEA are near about 2.9%. So the
applied numerical and FEA methods converge well with the experimental works.
The novel damage detection procedures have been also developed for finding out the
faults in the structures. The fault identification methods are in the domain of neural
networks and statistical process control environments.
In Chapter-5, the rule- based recurrent neural networks (RNNs) methods are developed to
identify the faults in the damaged structural systems. The knowledge based Jordan’s
recurrent neural networks (JRNNs), Elman’s recurrent neural network (ERNNs) and the
combined approach of JRNNs, and ERNNs are mainly focused in this Chapter. The
Levenberg-Merquardt’s back propagation algorithm has been implemented to train the
proposed RNNs model with sum squared error. The different steps involved for organising
the training procedures of the proposed network using the Levenberg-Merquardt’s
algorithm are also explained. The architecture of the modified JRNNs, ERNNs and the
combined approach of JRNNs, and ERNNs are represented in Figures 5.6, 5.7 and 5.8
respectively. The detailed training procedures of the RNNs are already explained in
Chapter 9 Results and Discussion
148
Chapter-5. 600 numbers of training patterns are generated for each of the damaged
structure and the networks are trained with 1000 numbers of iterations. The relative crack
locations and depth of each multi-cracked structure are predicted using the RNNs
techniques. The comparisons of results among experimental, JRNNs, ERNNs and
combined approach of JRNNs, and ERNNs for relative crack locations are represented in
Tables 5.2(a), 5.3(a) and 5.4(a) for cantilever, simply supported and fixed-fixed structures
respectively, while those for relative crack depth are expressed in Tables 5.2(b), 5.3(b) and
5.4(b) respectively. It has been remarked that the disparities of results between
experiments and JRNNs are an average near about 6.5%, while those with ERNNs and the
combined approach of JRNNs, and ERNNs are about 5.5% and 4.5% respectively.
In Chapter-6, the innovative damage identification process has been evolved in the domain
of statistical process control method using the time series analysis concepts. The theory of
control chart analysis is established to know the existence of cracks in the structures. The
concepts of autoregressive (AR) process are applied to identify the faults in structures. In
AR process, the training patterns are generated by considering both the damaged and
undamaged states of the structures. The complete training and monitoring process of the
AR model are previously explicated in Chapter-6. The coefficients of AR model represent
the damage sensitive features of the structures. The displacement-time data are considered
as the input parameters to the AR model. Using the hypothesis of Fisher’s criterion, the
locations of the damages have been identified. The fourth order statistical moment method
for stiffness vector and probability density function are established to find out the relative
crack depth. The comparisons of results among FEA, theoretical and SPCM methods for
relative crack locations are represented in Tables 6.1(a), 6.2(a) and 6.3(a) for cantilever,
simply supported and fixed-fixed beams respectively, while those for relative crack depth
are expressed in Tables 6.1(b), 6.2(b) and 6.3(b) respectively. It has been observed that the
variation of results between FEA and SPCM are near about 1.95%, while those between
experiments and SPCM are of 3.9% approximately.
In order to improve the results obtained from Chapters 5 and 6, a combined hybrid neuro-
autoregressive method has been elaborated in Chapter-7. The outputs from the AR model
are fed to hybridised JRNNs and ERNNs model as input to the newly developed network
model. The newly developed RNNs model has some additional inputs (outputs from AR
model) along with their usual inputs. The training and monitoring procedures of newly
developed model are already explained in Chapters-5 and 6. The comparison of results
Chapter 9 Results and Discussion
149
between combined hybrid neuro-autoregressive process and experiments are average
errors of 3.5%, while those with FEA are of 1.55% approximately. The comparison of
different damage detection methods are shown in Figure 9.1. It has been concluded that
the hybrid-neuro AR process produce better results.
9.3 Summary
Several intelligent methods such as JRNNs, ERNNs, hybridised approach of JRNNs and
ERNNs, SPC method, and the integrated approach of combined hybrid neuro-
autoregressive methods have been implemented for fault detection in multi cracked
structures under moving mass. The accuracy and exactness of each intelligent method
have been compared with experiments, FEA and theory. It has been concluded that the
combined hybrid neuro-autoregressive method yields better results as comparison to other
results.
0
1
2
3
4
5
6
7
8
1 2 3 4 5 6 7 8 9 10
Perc
en
ata
ge v
alu
e
Number of observations
JRNNs
ERNNs
Combined JRNNs and
ERNNsAR process
Hybrid neuro-AR process
Theory
FEA
Figure 9.1 Comparison of results among different damage
detection methods
150
Chapter 10
CONCLUSIONS AND SUGGESTION
FOR FURTHER RESEARCH
10.1 Introduction
In the present analysis, the vibration analyses of damaged structures subjected to transit
mass are studied. The responses of the multi-cracked structures with different boundary
conditions have been analyzed with different methods. The damage identification
procedures have been carried out using different soft computing approaches.
10.2 Contributions
The dynamic responses of beam type’s structures with multiple cracks subjected to
moving object are studied. The theoretical-numerical solutions of the beam structures
under transit mass has been formulated and consequently solved by applying Runge-Kutta
fourth order integration scheme. The FEA and experimental analysis have been carried out
to verify the applied numerical method. The FEA method (The full method transient
dynamic analysis) in ANSYS WORKBENCH 2015 domain has been implemented to find
out the responses of the structures subjected to transit mass. The laboratory tests for each
of the structural systems have been performed to check the accuracy of the proposed
computational and FEA methods. The responses of the structures subjected to transit mass
are determined with various damage configurations, moving speed and traversing mass of
the structural systems. The influences of different parameters like traversing mass, moving
speed, crack locations and depth on the response of the structures are also investigated.
The various intelligent methods such as RNNs, SPC, and integrated approach of RNNs
and SPC based methods have been applied as inverse problems for fault detection in the
damaged structures. Some knowledge based RNNs like JRNNs, ERNNs, and hybrid
approach of JRNNs and ERNNs are implemented to estimate the relative crack locations
and depth. The Levenberg-Merquardt’s algorithm has been implemented to train the
networks. The autoregressive (AR) process in SPC domain are applied for the prediction
Chapter 10 Conclusions and Suggestion for Further Research
151
of relative crack locations. The fourth order statistical moment with central differential
scheme is applied to quantify the crack severities. For the improvement of the above
methods and results, the RNNs and SPC methods are integrated to form a combine hybrid
neuro-autoregressive method. This hybrid neuro-autoregressive method has been used for
the prediction of relative crack locations and depth in the structures.
10.3 Conclusions
The dynamic analysis and fault detection of cracked structures under moving mass has
been studied. The influences of different parameters such as traversing mass, moving
speed, crack locations and depth on the dynamic responses of the structures under moving
mass have been investigated. The FEA and experimental verifications have been carried
out to check the accuracy and exactness of the proposed computational method (Runge-
Kutta method). It has been observed that the results obtained from the numerical analysis
are varying with error about 5% with experimental procedures and about 2% with FEA.
So the implemented computational method agrees well with FEA and experiments.
The relative crack locations and depth have been predicted using some knowledge based
intelligent methods like RNNs, statistical process controls (SPC), and the integrated
approach of the RNNs and SPC. The errors of result between experiments and JRNNs is
near about 6.5%, while those with ERNNs and the hybrid approach of JRNNs and ERNNs
are about of 5.5% and 4.4% respectively. The damage detection procedures have been
carried out using the concept of time series analysis in SPC domain. The autoregressive
(AR) process in SPC domain are applied for the prediction of relative crack locations. The
fourth order statistical moment with central differential scheme is applied to quantify the
crack severities. The disparities of results between the AR process and experiments are
with an average error of 3.95%, while that with FEA is 1.95% approximately. The results
obtained from the hybrid neuro-autoregressive method are converged with an error of
3.5% with experiments and 1.55% with FEA respectively.
It has been come to an end that the combined hybrid neuro-autoregressive gives the best
result as compared to other methods addressed in the current analogy. Using the neural
network and SPC controller, the online fault detection can be carried out for beam
structures under traversing mass.
Chapter 10 Conclusions and Suggestion for Further Research
152
10.4 Recommendation for future study
The problem may be extended in the following ways:
The theoretical-numerical solution of damaged structure subjected to moving mass
can be formulated under different foundation of structures.
The vibration analysis of damaged shaft under traversing mass is to be carried out.
The response analysis of structure subjected to transit mass at variable speed can
be found out.
The damage detection procedures can be developed using the integrated approach
of RNNs and Genetic algorithm, Statistical process control methods such as
autoregressive moving approach, Statistical moment based approach and some
natures inspired algorithm like multi Swarm Fruit Fly, Collaborative-Climb
Monkey and Climbing Hill algorithms.
153
Appendix
Appendix-A
Crack analysis on the vibration characteristics of cantilever beam:
A cracked cantilever beam (Fig A-1.) of length ‘L’, width ‘B’, thickness ‘H’ with multiple
cracks of crack depth ‘ 1,2,3d ’, at distance of ‘ 1,2,3L ’ from the fixed end is considered for the
analysis. The analysis of the cracked beam is carried out as follows
( , )u x t = Longitudinal vibration displacement functions of the crack section
( , )y x t = Transverse vibration displacement functions of the crack section.
One can define the normalized function for the cracked structure with multiple cracks in
normalized form as
1 1 2( ) cos( ) sin( )u uu x A K x A K x (A-1)
2 3 4( ) cos( ) sin( )u uu x A K x A K x (A-2)
3 5 6( ) cos( ) sin( )u uu x A K x A K x (A-3)
4 7 8( ) cos( ) sin( )u uu x A K x A K x (A-4)
1 9 10 11 12( ) cosh( ) sinh( ) cos( ) sin( )y y y yy x A K x A K x A K x A K x (A-5)
2 13 14 15 16( ) cosh( ) sinh( ) cos( ) sin( )y y y yy x A K x A K x A K x A K x (A-6)
3 17 18 19 20( ) cosh( ) sinh( ) cos( ) sin( )y y y yy x A K x A K x A K x A K x (A-7)
4 21 22 23 24( ) cosh( ) sinh( ) cos( ) sin( )y y y yy x A K x A K x A K x A K x (A-8)
The normalized form may be expressed as
1,2,3 1,2,3 /L L =Relative location of cracks.
/ , / , / , /x x L u u L y y L t t L
11 122 2
/ , , , , u u y yy
LE EIK L C C K C m AC m
The different values of ‘A’ ( ( 1,24))iA i can be determined from the different end
conditions.
The different end conditions of the cantilever beam are as follows
1 1 1(0) 0, (0) 0, (0) 0u y y (A-9)
Appendix
154
At the free end , then 1xx L xL
4 4 4(1) 0, (1) 0, (1) 0u y y (A-10)
At the crack segment
1 1 2 1 2 2 3 2 3 3 4 3( ) ( ), ( ) ( ), ( ) ( )u u u u u u (A-11)
1 1 2 1 2 2 3 2 3 3 4 3( ) ( ), ( ) ( ), ( ) ( )y y y y y y (A-12)
1 1 2 1 2 2 3 2 3 3 4 3( ) ( ), ( ) ( ), ( ) ( )y y y y y y (A-13)
1 1 2 1 2 2 3 2 3 3 4 3( ) ( ), ( ) ( ), ( ) ( )y y y y y y (A-14)
At the first crack segment also
1 11 1 2 1 1 111 2 1 1 1 12
( ) ( ) ( )( ) ( )
du L dy L dy LAE K u L u L K
dx dx dx
(A-15)
Similarly for the second and third crack sections, one can express
2 2 3 22 2 2 211 3 2 2 2 12
( )( ) ( )( ) ( )
dy Ldu L dy LAE K u L u L K
dx dx dx
(A-16)
3 33 3 4 3 3 311 4 3 3 3 12
( ) ( ) ( )( ) ( )
du L dy L dy LAE K u L u L K
dx dx dx
(A-17)
Equation (A-15) is for discontinuity due to axial deformation before and after the first
crack.
Similarly due to the discontinuity of slope to the before and after of the first crack section,
one can express
21 11 1 2 1 1 1
21 2 1 1 1 222
( ) ( ) ( )[ ( ) ( )]
d y L dy L dy LEI K u L u L K
dx dx dx
(A-18)
Similarly for the second and third crack sections, one can express
22 2 3 22 2 2 2
21 3 2 2 2 222
( )( ) ( )[ ( ) ( )]
dy Ld y L dy LEI K u L u L K
dx dx dx
(A-19)
23 33 3 4 3 3 3
21 4 3 3 3 222
( ) ( ) ( )[ ( ) ( )]
d y L dy L dy LEI K u L u L K
dx dx dx
(A-20)
By inversing the compliance matrix, the local flexibility matrix element ‘ ijK ’ may be
determined.
One can determine the compliance matrix element by considering the strain energy at the
crack location- 12
0
( )
d
ij
i j
C J a ddP P
(A-21)
Appendix
155
Where ( )J a = Strain energy density function
1iP =Axial force, 2iP =Bending moment
156
Bibliography
1. N. Sridharan and A. K. Mallik, “Numerical analysis of vibration of beams
subjected to moving loads”, Journal of Sound and Vibration, vol. 65, no. l, pp.
147-150, 1979.
2. S. A. Q. Siddiqui, M. F. Golnaraghi and G. R. Heppler, “Dynamics of a flexible
cantilever beam carrying a moving mass”, Nonlinear Dynamics, vol. 15, pp. 137-
154, 1998.
3. J. E. Akin and M. Mofld, “Numerical solution for response of beams with
moving mass”, Journal of Structural Engineering, vol. 115, pp. 120-131, 1989.
4. M. M. Stanisic and J. C. Hardin, “On the response of beams to an arbitrarily
number of concentrated moving masses”, Journal of the Franklin Institute,
vol. 287, no. 2, pp. 115-124, 1969.
5. M. Olsson, “On the fundamental moving load problem”, Journal of Sound and
Vibration, vol. 145, no. 2, pp. 299-307, 1991.
6. M. Moffid and J. E. Akin, “Discrete element response of beams with travelling
mass”, Advances in Engineering Structure, vol. 25, pp. 321-331, 1996.
7. H. C. Kwon, M. C. Kim and I. W. Lee, “Vibration control of bridges under
moving loads”, Computers & Structures, vol. 66, no. 4, pp. 473-480, 1998.
8. M. A. Mahmoud and M. A. Abouzaid, “Dynamic response of a beam with a
crack subject to a moving mass”, Journal of Sound and Vibration, vol. 256, no. 4,
pp. 591-603, 2002.
9. J. Li, M. Su and L. Fan, “Natural frequency of railway girder bridges under
vehicle loads”, Journal of Bridge Engineering, vol. 8, no. 4, pp. 199-203, 2003.
10. C. Bilello, L. A. Bergman and D. Kuchma, “Experimental investigation of a
small-scale bridge model under a moving mass”, Journal of Structural
Engineering, vol. 130, no. 5, pp. 799-804, 2004.
11. C. Bilello, L. A. Bergman, “Vibration of damaged beams under a moving mass:
theory and experimental validation”, Journal of Sound and Vibration, vol. 274,
pp. 567-582, 2004.
12. Y. B. Yang, C. W. Lin and J. D. Yau, “Extracting bridge frequencies from the
dynamic response of a passing vehicle”, Journal of Sound and Vibration, vol.
272, pp. 471-493, 2004.
13. M. Majka and M. Hartnett, “Effects of speed, load and damping on the dynamic
response of railway bridges and vehicles”, Computers and Structures, vol. 86, pp.
556-572, 2008.
14. A. Garinei and G. Risitano, “Vibrations of railway bridges for high speed trains
under moving loads varying in time”, Engineering Structures, vol. 30, pp. 724-
732, 2008.
Reference
157
15. Y. B. Yang and K. C. Chang, “Extracting the bridge frequencies indirectly from a
passing vehicle: Parametric study”, Engineering Structures, vol. 31, pp. 2448-
2459, 2009.
16. M. Dehestani, M. Mofid and A. Vafai, “Investigation of critical influential speed
for moving mass problems on beams”, Applied Mathematical Modelling, vol. 33,
pp. 3885-3895, 2009.
17. K. Liu, G. D. Roeck and G. Lombaert, “The effect of dynamic train–bridge
interaction on the bridge response during a train passage”, Journal of Sound and
Vibration, vol. 325, pp. 240-251, 2009.
18. D. M. Siringoringo and Y. Fujino, “Estimating bridge fundamental frequency
from vibration response of instrumented passing vehicle: Analytical and
experimental study”, Advances in Structural Engineering, vol. 15, no. 3, pp. 443-
460, 2012.
19. H. Xia, N. Zhang and W. W. Guo, “Analysis of resonance mechanism and
conditions of train-bridge system”, Journal of Sound and Vibration, vol. 297, pp.
810-822, 2006.
20. M. Majkaa and M. Hartnett, “Dynamic response of bridges to moving trains: A
study on effects of random track irregularities and bridge skewness”, Computers
and Structures, vol. 87, pp. 1233-1252, 2009.
21. S. -H. Ju, H. -T. Lin and J. -Y. Huang, “Dominant frequencies of train-induced
vibrations”, Journal of Sound and Vibration, vol. 319, pp. 247-259, 2009.
22. H. -I. Yoon, I. -S. Son and S. -J. Ahn, “Free vibration analysis of Euler-Bernoulli
beam with double cracks”, Journal of Mechanical Science and Technology, vol.
21, pp. 476-485, 2007.
23. M. A. Mahmoud, “Stress intensity factors for single and double edge cracks in a
simple beam subject to a moving load”, International Journal of Fracture, vol.
111, pp. 151-161, 2001.
24. G. Michaltsos, D. Sophianopoulos and A. N. Kounadis, “The effect of a moving
mass and other parameters on the dynamic response of a simply supported
beam”, Journal of Sound and Vibration, vol. 191, no. 3, pp. 357-362, 1996.
25. M. A. Mahmoud, “Effect of cracks on the dynamic response of a simple beam
subject to a moving load”, Journal of Rail and Rapid Transit, vol. 215, pp. 206-
215, 2001.
26. H. -P. Lin and S. -C. Chang, “Forced responses of cracked cantilever beams
subjected to a concentrated moving load”, International Journal of Mechanical
Sciences, vol. 48, pp. 1456-1463, 2006.
27. Ouyang, H., “Moving-load dynamic problems: A tutorial (with a brief
overview)”, Mechanical Systems and Signal Processing, 2011, 25, 2039-2060.
Reference
158
28. Ariaei, A., Ziaei-Rad, S. and Ghayour, M, “Vibration analysis of beams with
open and breathing cracks subjected to moving masses”, Journal of Sound and
Vibration, 2009, 326, 709-724.
29. S. E. Azam, M. Mofid and R. A. Khoraskani, “Dynamic response of Timoshenko
beam under moving mass”, Scientia Iranica A, vol. 20, no. 1, pp. 50-56, 2013.
30. G. T. Michaltsos and A. N. Kounadis, “The effects of centripetal and Coriolis
forces on the dynamic response of light bridges under moving loads”, Journal of
Vibration and Control, vol. 7, pp. 315-326, 2001.
31. Y. Pala and M. Reis, “Dynamic response of a cracked beam under a moving mass
load”, Journal of Engineering Mechanics, vol. 129, pp. 1229-1238, 2013.
32. M. Reis and Y. Pala, “Vibration of a cracked cantilever beam under moving mass
load”, Journal of Civil Engineering and Management, vol. 18, no. 1, pp. 106-113,
2012.
33. X. Shi, C. S. Cai and S. Chen, “Vehicle induced dynamic behavior of short-span
slab bridges considering effect of approach slab condition”, Journal of Bridge
Engineering, vol. 13, pp. 83-92, 2008.
34. E. Esmailzadeh and M. Ghorashi, “Vibration analysis of a Timoshenko beam
subjected to a travelling mass”, Journal of Sound and Vibration, vol. 199, no. 4,
pp. 615-628, 1997.
35. H. P. Lee, “The dynamic response of a Timoshenko beam subjected to a moving
mass”, Journal of Sound and Vibration, vol. 198, no. 2, pp. 249-256, 1996.
36. E. Khalily, M. E. Golnaraghi and G. R. Heppler, “On the dynamic behaviour of a
flexible beam carrying a moving mass”, Nonlinear Dynamics, vol. 5, pp. 493-
513, 1994.
37. M. Mofid and M. Shadnam, “On the response of beams with internal hinges,
under moving mass”, Advances in Engineering Software, vol. 31, pp. 323-328,
2000.
38. A. O. Cifuentes, “Dynamic response of a beam excited by a moving mass”, Finite
Elements in Analysis and Design, vol. 5, pp. 237-24, 1989.
39. A. Nikkhoo, F. R. Rofooei and M. R. Shadnam, “Dynamic behavior and modal
control of beams under moving mass”, Journal of Sound and Vibration, vol. 306,
pp. 712-724, 2007.
40. D. A. Grant, “The effect of rotary inertia and shear deformation on the frequency
and normal mode equations of uniform beams carrying a concentrated mass”,
Journal of Sound and Vibration, vol. 57, no. 3, pp. 357-365, 1978.
41. D. Thambiratnam and Y. Zhuge, “Dynamic analysis of beams on an elastic
foundation subjected to moving loads”, Journal of Sound and Vibration, vol. 198,
no. 2, pp. 149-169, 1996.
42. G. V. Rao, “Linear dynamics of an elastic beam under moving loads”, Journal of
Vibration and Acoustics, vol. 122, pp. 281-289, 2000.
Reference
159
43. A. Yavari, M. Nouri and M. Mofid, “Discrete element analysis of dynamic
response of Timoshenko beam under moving mass”, Advances in Engineering,
vol. 33, pp. 143-153, 2002.
44. M. Abu-Hilal, “Dynamic response of a double Euler–Bernoulli beam due to a
moving constant load”, Journal of Sound and Vibration, vol. 297, pp. 477-491,
2006.
45. H. P. Wang and J. L. K. Zhang, “Vibration analysis of the maglev guideway with
the moving load”, Journal of Sound and Vibration, vol. 305, pp. 621-640, 2007.
46. J. Yang, Y. Chen, Y. Xiang and X. L. Jia, “Free and forced vibration of cracked
inhomogeneous beams under an axial force and a moving load”, Journal of Sound
and Vibration, vol. 312, pp. 166-181, 2008.
47. T. Yan, S. Kitipornchai, J. Yang and X. Q. He, “Dynamic behaviour of edge-
cracked shear deformable functionally graded beams on an elastic foundation
under a moving load”, Composite Structures, vol. 93, pp. 2992-3001, 2011.
48. M. Shafiei and N. Khaji, “Analytical solutions for free and forced vibrations of a
multiple cracked Timoshenko beam subject to a concentrated moving load”, Acta
Mechanica, vol. 221, pp. 79-97, 2011.
49. S. -I. Suzuki, “Dynamic behaviour of a finite beam subjected to travelling
loads with acceleration”, Journal of Sound and Vibration, vol. 55, no. l, pp.
65-70, 1977.
50. T. R. Hamada, “Dynamic analysis of a beam under a moving force: a
double Laplace transform solution”, Journal of Sound and Vibration, vol. 74,
no. 2, pp. 221-233, 1981.
51. H. P. Lee and T. Y. Ng, “Dynamic response of a cracked beam subject to a
moving load”, Acta Mechanica, vol. 106, pp. 221-230, 1994.
52. M. Ichikawa, Y. Miyakawa and A. Matsuda, “Vibration analysis of the
continuous beam subjected to a moving mass”, Journal of Sound and Vibration,
vol. 230, no. 3, pp. 493-506, 2009.
53. J. J. Wu, A. R. Whittaker and M. P. Cartmell, “Dynamic responses of structures
to moving bodies using combined finite element and analytical methods”,
International Journal of Mechanical Sciences, vol. 43, pp. 2555-2579, 2001.
54. A. K. Mallik, S. Chandra and A. B. Singh, “Steady-state response of an
elastically supported infinite beam to a moving load”, Journal of Sound and
Vibration, vol. 291, pp. 1148-1169, 2006.
55. K. Aydin, “Vibratory characteristics of Euler-Bernoulli beams with an arbitrary
number of cracks subjected to axial load”, Journal of Vibration and Control, vol.
14, no. 4, pp. 485-510, 2008.
56. R. Sieniawska, P. Sniady and S. Zukowski, “Identification of the structure
parameters applying a moving load”, Journal of Sound and Vibration, vol. 319,
pp. 355-365, 2009.
Reference
160
57. T. Yan and J. Yang, “Forced vibration of edge-cracked functionally graded
beams due to a transverse moving load”, Procedia Engineering, vol. 14, pp.
3293-3300, 2011.
58. G. Chen, L. Qian and Q. Yin, “Dynamic analysis of a Timoshenko beam
subjected to an accelerating mass using spectral element method”, Shock and
Vibration, 2014, Article ID 768209.
59. R. Zarfam and A. R. Khaloo, “Vibration control of beams on elastic foundation
under a moving vehicle and random lateral excitations”, Journal of Sound and
Vibration, vol. 331, pp. 1217-1232, 2012.
60. C. Johansson, C. Pacoste and R. Karoumi, “Closed-form solution for the mode
superposition analysis of the vibration in multi-span beam bridges caused by
concentrated moving loads”, Computers and Structures, vol. 119, pp. 85-94,
2013.
61. P. Lou and F. T. K. Au, “Finite element formulae for internal forces of Bernoulli-
Euler beams under moving vehicles”, Journal of Sound and Vibration, vol. 332,
pp. 1533-1552, 2013.
62. P. Museros, E. Moliner and M. D. Martinez-Rodrigo, “Free vibrations of simply-
supported beam bridges under moving loads: Maximum resonance, cancellation
and resonant vertical acceleration”, Journal of Sound and Vibration, vol. 332, pp.
326-345, 2013.
63. R. Zarfam, A. R. Khaloo and A. Nikkhoo, “On the response spectrum of
Euler–Bernoulli beams with a moving mass and horizontal support
excitation”, Mechanics Research Communications, vol. 47, pp. 77-83, 2013.
64. H. Azimi, K. Galal and O. A. Pekau, “A numerical element for vehicle–bridge
interaction analysis of vehicles experiencing sudden deceleration”, Engineering
Structures, vol. 49, pp. 792-805, 2013.
65. A. Cicirello and A. Palmeri, “Static analysis of Euler–Bernoulli beams with
multiple unilateral cracks under combined axial and transverse loads”,
International Journal of Solids and Structures, vol. 51, pp. 1020-1029, 2014.
66. H. Zhong, M. Yang and Z. Gao, “Dynamic responses of prestressed bridge and
vehicle through bridge-vehicle interaction analysis”, Engineering Structures, vol.
87, pp. 116-125, 2015.
67. P. A. Costa, A. Colaco, R. Calcada and A. S. Cardoso, “Critical speed of railway
tracks. Detailed and simplified approaches”, Transportation Geotechnics, vol. 2,
pp. 30-46, 2015.
68. C. Fu, “The effect of switching cracks on the vibration of a continuous beam
bridge subjected to moving vehicles”, Journal of Sound and Vibration, vol. 339,
pp. 157-175, 2015.
Reference
161
69. C. Fu, “Dynamic behavior of a simply supported bridge with a switching crack
subjected to seismic excitations and moving trains”, Engineering Structures, vol.
110, pp. 59-69, 2016.
70. H. Aied, A. González and D. Cantero, “Identification of sudden stiffness changes
in the acceleration response of a bridge to moving loads using ensemble empirical
mode decomposition”, Mechanical Systems and Signal Processing, vol. 66-67,
pp. 314-338, 2016.
71. J. Hino, T. Yoshimura, K. Konishi and N. Ananthanarayana, “A finite element
method vibration of a bridge prediction of the subjected to a moving vehicle
load”, Journal of Sound and Vibration, vol. 96, no. l, pp. 45-53, 1984.
72. G. R. Bhashyam and G. Prathap, “Galerkin finite element method for non-
linear beam vibrations”, Journal of Sound and Vibration, vol. 72, no. 2, pp.
191-203, 1980.
73. M. Olsson, “Finite element, modal co-ordinate analysis of structures subjected to
moving loads”, Journal of Sound and Vibration, vol. 99, no. l, pp. l-12, 1985.
74. T. Yoshimura, J. Hino, T. Kamata and N. Ananthanarayana, “Random vibration
of a non-linear beam subjected to a moving load: A finite element method
analysis,” Journal of Sound and Vibration, vol. 122, no. 2, pp. 317-329, 1988.
75. Y. H. Lin and M. W. Trethewey, “Finite element analysis of elastic beams
subjected to moving dynamic loads”, Journal of Sound and Vibration, vol. 136,
no. 2, pp. 323-342, 1990.
76. 76. T. P. Chang and Y. -N. Liu, “Dynamic finite element analysis of a
nonlinear beam subjected to a moving load”, International Journal of Solids
Structures, vol. 33, no. 12, pp. 1673-1688, 1996.
77. J. R. R1eker and M. W. Trethewey, “Finite element analysis of an elastic beam
structure subjected to a moving distributed mass train”, Mechanical Systems and
Signal Processing, vol. 13, no. 1, pp. 31-51, 1999.
78. M. -K. Songa, H. -C. Noh and C. -K. Choi, “A new three-dimensional finite
element analysis model of high-speed train-bridge interactions”, Engineering
Structures, vol. 25, pp. 1611-1626, 2003.
79. S. S. Law and X. Q. Zhu, “Dynamic behavior of damaged concrete bridge
structures under moving vehicular loads”, Engineering Structures, vol. 26, pp.
1279-1293, 2004.
80. Y. B. Yang and C. W. Lin, “Vehicle–bridge interaction dynamics and potential
applications”, Journal of Sound and Vibration, vol. 284, pp. 205-226, 2005.
81. L. Kwasniewski, H. Li, J. Wekezer and J. Malachowsk, “Finite element analysis
of vehicle-bridge interaction”, Finite Elements in Analysis and Design, vol. 42,
pp. 950-959, 2006.
Reference
162
82. S. -H. Ju and H. -T. Lin, “A finite element model of vehicle–bridge interaction
considering braking and acceleration”, Journal of Sound and Vibration, vol. 303,
pp. 46-57, 2007.
83. C. I. Bajer and B. Dyniewicz, “Virtual functions of the space–time finite element
method in moving mass problems”, Computers and Structures, vol. 87, pp. 444-
455, 2009.
84. V. Kahya, “Dynamic analysis of laminated composite beams under moving loads
using finite element method”, Nuclear Engineering and Design, vol. 243, pp. 41-
48, 2012.
85. B. Dyniewicz, “Space-time finite element approach to general description of a
moving inertial load”, Finite Elements in Analysis and Design, vol. 62, pp. 8-17,
2012.
86. I. Esen, “A new finite element for transverse vibration of rectangular thin plates
under a moving mass”, Finite Elements in Analysis and Design, vol. 66, pp. 26-
35, 2013.
87. W. Zhang, C. S. Cai and F. Pan, “Finite element modeling of bridges with
equivalent orthotropic material method for multi-scale dynamic loads”,
Engineering Structures, vol. 54, pp. 82-93, 2013.
88. J. V. Amiri, A. Nikkhoo, M. R. Davoodi and M. E. Hassanabadi, “Vibration
analysis of a Mindlin elastic plate under a moving mass excitation by
eigenfunction expansion method”, Thin-Walled Structures, vol. 62, pp. 53-64,
2013.
89. S. H. Ju, “Improvement of bridge structures to increase the safety of moving
trains during earthquakes”, Engineering Structures, vol. 56, pp. 501-508, 2013.
90. F. B. Nejad, A. Khorram and M. Rezaeian, “Analytical estimation of natural
frequencies and mode shapes of a beam having two cracks”, International
Journal of Mechanical Sciences, vol. 78, pp. 193-202, 2014.
91. H. Aied and A. Gonzalez, “Theoretical response of a simply supported beam with
a strain rate dependant modulus to a moving load”, Engineering Structures, vol.
77, pp. 95-108, 2014.
92. J. Wu, J. Liaug and S. Adhikar, “Dynamic response of concrete pavement
structure with asphalt isolating layer under moving loads”, Journal of Traffic and
Transportation Engineering, vol. 1, no. 6, pp. 439-447, 2014.
93. P. C. Jorge, F. M. F. Simoes and A. P. Costa, “Dynamics of beams on non-
uniform nonlinear foundations subjected to moving loads”, Computers and
Structures, vol. 148, pp. 26-34, 2015.
94. P. C. Jorge, F. M. F. Simoes and A. P. Costa, “Finite element dynamic analysis of
finite beams on a bilinear foundation under a moving load”, Journal of Sound and
Vibration, vol. 346, pp. 328-344, 2015.
Reference
163
95. R. Alebrahim, S. M. Haris, N. A. N. Mohamed and S. Abdullah, “Vibration
analysis of self-healing hybrid composite beam under moving mass”, Composite
Structures, vol. 119, pp. 463-476, 2015.
96. H. Ozturk, Z. Kiral and B. G. Kiral, “Dynamic analysis of elastically supported
cracked beam subjected to a concentrated moving load”, Latin American Journal
of Solids and Structures”, vol. 13, no. 1, pp. 175-200, 2016.
97. C. Fu, “Dynamic behavior of a simply supported bridge with a switching crack
subjected to seismic excitations and moving trains”, Engineering Structures, vol.
110, pp. 59-69, 2016.
98. N. D. Beskou, S. V. Tsinopoulos and D. D. Theodorakopoulos, “Dynamic elastic
analysis of 3-Dflexible pavements under moving vehicles: A unified FEM
treatment”, Soil Dynamics and Earthquake Engineering, vol. 82, pp. 63-72, 2016.
99. A. Dutta and S. Talukdar, “Damage detection in bridges using accurate modal
parameters”, Finite Elements in Analysis and Design, vol. 40, pp. 287-304, 2004.
100. M. I. Friswell and J. E. T. Penny, “Crack modeling for structural health
monitoring”, Structural health monitoring, vol. 1, no. 2, pp. 139-148, 2002.
101. J. W. Lee, A. D. Kim, C. B. Yun, J. H. Yi and J. M. Shim, “Health-monitoring
method for bridges under ordinary traffic loadings”, Journal of Sound and
Vibration, vol. 257, no. 2, pp. 247-264, 2002.
102. M. A. B. Abdo and M. Hori, “A numerical study of structural damage detection
using changes in the rotation of mode shapes”, Journal of Sound and Vibration,
vol. 251, no. 2, pp. 227-239, 2002.
103. J. T. Kim and N. Stubbs, “Improved damage identification method based on
modal information”, Journal of Sound and Vibration, vol. 252, no. 2, pp. 223-
238, 2002.
104. T. G. Chondros, A. D. Dimarogonas and J. Yao, “A continuous cracked beam
vibration theory”, Journal of Sound and Vibration, vol. 215, no. 1, pp. 17-34,
1998.
105. T. G. Chondros and A. D. Dimarogonas, “Vibration of a cracked cantilever
beam”, Journal of Vibration and Acoustics, vol. 120, pp. 742-746, 1998.
106. F. Khoshnoudian and A. Eafandiari, “Structural damage diagnosis using modal
data”, Scientia Iranica A, vol. 18, no. 4, pp. 853-860, 2011.
107. A. S. J. Swamidas, X. Yang and R. Seshadri, “Identification of cracking in beam
structures using Timoshenko and Euler formulations”, Journal of Engineering
Mechanics, vol. 130, pp. 1297-1308, 2004.
108. D. F. Mazurek and J. T. DeWolf, “Experimental study of bridge monitoring
technique”, Journal of Structural Engineering, vol. 116, pp. 2532-2549, 1990.
109. R. Ruotolo and C. Surace, “Damage assessment of multiple cracked beams:
Numerical results and experimental validation”, Journal of Sound and Vibration,
vol. 206, no. 4, pp. 567-588, 1997.
Reference
164
110. A. K. Pandey and M. Biswas, “Damage detection in structures using the changes
in flexibility”, Journal of Sound and Vibration, vol. 169, no. 1, pp. 3-17, 1994.
111. T. D. Chaudhari and S. K. Maiti, “A study of vibration of geometrically
segmented beams with and without crack”, International Journal of Solids and
Structures, vol. 37, pp. 761-779, 2000.
112. L. Majumder and C. S. Manohar, “A time-domain approach for damage detection
in beam structures using vibration data with a moving oscillator as an excitation
source”, Journal of Sound and Vibration, vol. 268, pp. 699-716, 2003.
113. S. Chinchalkar, “Determination of crack location in beams using natural
frequencies”, Journal of Sound and Vibration, vol. 247, no. 3, pp. 417-429, 2001.
114. N. Haritos and J. S. Owen, “The use of vibration data for damage detection in
bridges: A comparison of system identification and pattern recognition
approaches”, Structural Health Monitoring, vol. 3, no. 2, pp. 141-163, 2004.
115. H. Nahvi and M. Jabbari, “Crack detection in beams using experimental modal
data and finite element model”, International Journal of Mechanical Sciences,
vol. 47, pp. 1477-1497, 2005.
116. A. Alvandi and C. Cremona, “Assessment of vibration-based damage
identification techniques”, Journal of Sound and Vibration, vol. 292, pp. 179-
202, 2006.
117. S. Wang, Q. Ren and P. Qiao, “Structural damage detection using local damage
factor”, Journal of Vibration and Control, vol. 12, no. 9, pp. 955-973, 2006.
118. L. Fushun, L. Huajun, Y. Guangming, Z. Yantao, W. Weiying and S. Wanqing,
“New damage-locating method for bridges subjected to a moving load”, Journal
of Ocean University of China, vol. 6, no. 2, pp. 199-204, 2007.
119. L. Yu, H. Tommy and T. Chan, “Recent research on identification of moving
loads on bridges”, Journal of Sound and Vibration, vol. 305, pp. 3-21, 2007.
120. A. S. Sekhar, “Multiple cracks effects and identification”, Mechanical Systems
and Signal Processing, vol. 22, pp. 845-878, 2008.
121. N. Khaji, M. Shafiei and M. Jalalpour, “Closed-form solutions for crack detection
problem of Timoshenko beams with various boundary conditions”, International
Journal of Mechanical Sciences, vol. 51, pp. 667-681, 2009.
122. J. Lee, “Identification of multiple cracks in a beam using vibration amplitudes”,
Journal of Sound and Vibration, vol. 326, pp. 205-212, 2009.
123. Z. Zhu, S. German and I. Brilakis, “Detection of large-scale concrete columns for
automated bridge inspection”, Automation in Construction, vol. 19, pp. 1047-
1055, 2010.
124. Q. Zhang, L. Jankowski and Z. Duan, “Simultaneous identification of moving
masses and structural damage”, Structural and Multidisciplinary Optimization,
vol. 42, pp. 907-922, 2010.
Reference
165
125. F. B. Sayyad and B. Kumar, “Identification of crack location and crack size in a
simply supported beam by measurement of natural frequencies”, Journal of
Vibration and Control, vol. 18, no. 2, pp. 183-190, 2010.
126. M. Dilena and A. Morassi, “Dynamic testing of a damaged bridge”, Mechanical
Systems and Signal Processing, vol. 25, pp. 1485-1507, 2011.
127. N. Roveri and A. Carcaterra, “Damage detection in structures under traveling
loads by Hilbert–Huang transform”, Mechanical Systems and Signal Processing,
vol. 28, pp. 128-144, 2012.
128. K. V. Nguyen, “Comparison studies of open and breathing crack detections of a
beam-like bridge subjected to a moving vehicle”, Engineering Structures, vol. 51,
pp. 306-314, 2013.
129. J. Li and S. S. Law, “Damage identification of a target substructure with moving
load excitation”, Mechanical Systems and Signal Processing, vol. 30, pp. 78-90,
2012.
130. S. Q. Zhu and S. S. Law, “Damage detection in simply supported concrete bridge
structure under moving vehicular loads”, Journal of Vibration and Acoustics, vol.
129, pp. 58-65, 2007.
131. N. T. Khiem and T. V. Lien, “Multi-crack detection for beam by the natural
frequencies”, Journal of Sound and Vibration, vol. 273, pp. 175-184, 2004.
132. D. P. Patil and S. K. Maiti, “Experimental verification of a method of detection of
multiple cracks in beams based on frequency measurements”, Journal of Sound
and Vibration, vol. 281, pp. 439-451, 2005.
133. V. Pakrashi, A. O’Connor and B. Basu, “A bridge-vehicle interaction based
experimental investigation of damage evolution”, Structural Health Monitoring,
vol. 9, no. 4, pp. 285-296, 2010.
134. Z. Zhou, B. Zhang, K. Xia, X. Li, G. Yan and K. Zhang, “Smart film for crack
monitoring of concrete bridges”, Structural Health Monitoring, vol. 10, no. 3, pp.
275-289, 2010.
135. A. S. Bouboulas and N. K. Anifantis, “Finite element modeling of a vibrating
beam with a breathing crack: observations on crack detection”, Structural Health
Monitoring, vol. 10, no. 2, pp. 131-145, 2011.
136. J. W. Zhan, H. Xia, S. Y. Chen and G. D. Roeck, “Structural damage
identification for railway bridges based on train-induced bridge responses and
sensitivity analysis”, Journal of Sound and Vibration, vol. 330, pp. 757-770,
2011.
137. A. Ghadami, A. Maghsoodi and H. R. Mirdamadi, “A new adaptable multiple-
crack detection algorithm in beam-like structures”, Archives of Mechanics, vol.
65, no. 6, pp. 469-483, 2013.
138. J. Hu and R. Y. Liang, “An integrated approach cracks using vibration
characteristics”, Journal of Franklin Institute, vol. 5, pp. 841-853, 1993.
Reference
166
139. B. Chomette, A. Fernandes and J. J. Sinou, “Cracks detection using active modal
damping and piezoelectric components”, Shock and Vibration, vol. 20, pp. 619-
631, 2013.
140. F. B. Nejad, A. Khorram and M. Rezaeian, “Analytical estimation of natural
frequencies and mode shapes of a beam having two cracks”, International
Journal of Mechanical Sciences, vol. 78, pp. 193-202, 2014.
141. D. Wang, W. Xiang and H. Zhu, “Damage identification in beam type structures
based on statistical moment using a two step method”, Journal of Sound and
Vibration, vol. 333, pp. 745-760, 2014.
142. K. V. Nguyen, “Mode shapes analysis of a cracked beam and its application for
crack detection”, Journal of Sound and Vibration, vol. 333, pp. 848-872, 2014.
143. L. -T. Lee and J. Y. Wu, “Identification of crack locations and depths in a multi-
cracks beam by using local adaptive differential quadrature method”, Journal of
Process Mechanical Engineering, pp. 1-13, 2014.
144. P. Nandakumar and K. Shankar, “Multiple crack damage detection of structures
using the two crack transfer matrix”, Structural Health Monitoring, vol. 13, no. 5,
pp. 548-561, 2014.
145. N. T. Khiem and H. T. Tran, “A procedure for multiple crack identification in
beam-like structures from natural vibration mode”, Journal of Vibration and
Control, vol. 20, no. 9, pp. 1417-1427, 2014.
146. M. M. Ettefagh, H. Akbari, K. Asadi and F. Abbasi, “New structural damage-
identification method using modal updating and model reduction”, Journal of
Mechanical Engineering Science, pp. 1-19, 2014.
147. C. Schallhorn and S. Rahmatalla, “Crack detection and health monitoring of
highway steel-girder bridges”, Structural Health Monitoring, vol. 14, no. 3, pp.
281-299, 2015.
148. W. -Y. He and S. Zhu, “Moving load-induced response of damaged beam and its
application in damage localization”, Journal of Vibration and Control, pp. 1-17,
2015.
149. H. Zhu, L. Mao and S. Weng, “Calculation of dynamic response sensitivity to
substructural damage identification under moving load”, Advances in Structural
Engineering, vol. 16, no. 9, pp. 1621-1632, 2013.
150. S. S. Law and Z. R. Lu, “Crack identification in beam from dynamic responses”,
Journal of Sound and Vibration, vol. 285, pp. 967-987, 2005.
151. S. S. Law and X. Q. Zhu, “Nonlinear characteristics of damaged concrete
structures under vehicular load”, Journal of Structural Engineering, vol. 131, no.
8, pp. 1277-1285, 2005.
152. J. Q. Bu, S. S. Law and X. Q. Zhu, “Innovative bridge condition assessment from
dynamic response of a passing vehicle”, Journal of Engineering Mechanics, vol.
132, pp. 1372-1379, 2006.
Reference
167
153. M. Karthikeyan, R. Tiwari and S. Talukdar, “Development of a technique to
locate and quantify a crack in a beam based on modal parameters”, Journal of
Vibration and Acoustics, vol. 129, pp. 390-395, 2007.
154. Z. Zhou, L. D. Wegner and B. F. Sparling, “Vibration-based detection of small-
scale damage on a bridge deck”, Journal of Structural Engineering, vol. 133, no.
9, pp. 1257-1267, 2007.
155. S. M. Al-said, “Crack detection in stepped beam carrying slowly moving mass”,
Journal of Vibration and Control, vol. 14, no. 12, pp. 1903-1920, 2008.
156. S. H. Yin and C. Y. Tang, “Identifying cable tension loss and deck damage in a
Cable-stayed bridge using a moving vehicle”, Journal of Vibration and Acoustics,
vol. 133, no. 2, 021007, 2011.
157. J. Li, S. S. Law and H. Hao, “Improved damage identification in bridge structures
subject to moving loads: Numerical and experimental studies”, International
Journal of Mechanical Sciences, vol. 74, pp. 99-111, 2013.
158. J. Li and H. Hao, “Damage detection of shear connectors under moving loads
with relative displacement measurements”, Mechanical Systems and Signal
Processing, vol. 60, pp. 124-150, 2015.
159. X. Kong, C. S. Cai and B. Kong, “Damage detection based on transmissibility of
a vehicle and bridge coupled system”, Journal of Engineering Mechanics, vol.
141, no. 1, p. 04014102, 2015.
160. K. Aydin, “Vibratory characteristics of Euler-Bernoulli beams with an arbitrary
number of cracks subjected to axial load”, Journal of Vibration and Control, vol.
14, no. 4, pp. 485-510, 2008.
161. M. A. Dar, N. Subramanian, A. R. Dar and J. Raju, “Experimental investigations
on the structural behaviour of a distressed bridge”, Structural Engineering and
Mechanics, vol. 56, no. 4, pp. 695-705, 2015.
162. Y. Oshima, K. Yamamoto and K. Sugiura, “ Damage assessment of a bridge
based on mode shapes estimated by responses of passing vehicles”, Smart
Structures and Systems, vol. 13, no. 5, pp. 731-753, 2014.
163. K. Mazanoglu, “A novel methodology using simplified approaches for
identification of cracks in beams” Latin American Journal of Solids and
Structures, vol. 12, pp. 2460-2479, 2015.
164. D. Feng and M. Feng, “Output-only damage detection using vehicle-induced
displacement response and mode shape curvature index”, Structural Control and
Health Monitoring, 2016, DOI: 10.1002/stc.
165. J. H. Chou and J. Ghaboussi, “Genetic algorithm in structural damage detection”,
Computers and Structures, vol. 79, pp. 1335-1357, 2001.
166. E. G. Shopova and N. G. V. Bancheva, “BASIC-A genetic algorithm for
engineering problems solution”, Computers and Chemical Engineering, vol. 30,
pp. 1293-1309, 2006.
Reference
168
167. X. He, M. Kawatani, T. Hayashikawa, H. Furuta and T. Matsumoto, “A bridge
damage detection approach using train-bridge interaction analysis and GA
optimization”, Procedia Engineering, vol. 14, pp. 769-776, 2011.
168. L. Sun, X. Cai and J. Yang, “Genetic algorithm-based optimum vehicle
suspension design using minimum dynamic pavement load as a design criterion”,
Journal of Sound and Vibration, vol. 301, pp. 18-27, 2007.
169. M. V. Baghmisheh, M. Peimani, M. H. Sadeghi and M. M. Ettefagh, “Crack
detection in beam-like structures using genetic algorithms”, Applied Soft
Computing, vol. 8, pp. 1150-1160, 2008.
170. V. Meruane and W. Heylen, “An hybrid real genetic algorithm to detect structural
damage using modal properties”, Mechanical Systems and Signal Processing,
vol. 25, pp. 1559-1573, 2011.
171. F. S. Buezas, M. B. Rosales and C. P. Filipich, “Damage detection with genetic
algorithms taking into account a crack contact model”, Engineering Fracture
Mechanics, vol. 78, pp. 695-712, 2011.
172. C. Na, S. P. Kim and H. G. Kwak, “Structural damage evaluation using genetic
algorithm”, Journal of Sound and Vibration, vol. 330, pp. 2772-2783, 2011.
173. M. Mehrjooa, N. Khaji and M. G. Ashtiany, “Application of genetic algorithm in
crack detection of beam-like structures using a new cracked Euler-Bernoulli
beam element”, Applied Soft Computing, vol. 13, pp. 867-880, 2013.
174. S. Y. Lee, “An advanced coupled genetic algorithm for identifying unknown
moving loads on bridge decks”, Mathematical Problems in Engineering, 2014,
Article ID 462341.
175. C. Chisari, C. Bedon and C. Amadio, “Dynamic and static identification of base-
isolated bridges using Genetic Algorithms”, Engineering Structures, vol. 102, pp.
80-92, 2015.
176. H. Alli, A. Ucar and Y. Demir, “The solutions of vibration control problems
using artificial neural networks”, Journal of the Franklin Institute, vol. 340, pp.
307-325, 2003.
177. X. Cao, Y. Sugiyama and Y. Mitsui, “Application of artificial neural networks to
load identification”, Computers and Structures, vol. 69, pp. 63-78, 1998.
178. M. A. Mahmoud and M. A. Kiefa, “Neural network solution of the inverse
vibration problem”, NDT&E International, vol. 32, pp. 91-99, 1999.
179. Z. Waszczyszyn and L. Ziemianski, “Neural networks in mechanics of structures
and materials-new results and prospects of application”, Computer and
Structures, vol. 79, pp. 2261-2276, 2001.
180. C. Zang and M. Imregun, “Structural damage detection using artificial neural
networks and measured FRF data reduced via principal component projection”,
Journal of Sound and Vibration, vol. 242, no. 5, pp. 813-827, 2001.
Reference
169
181. S. W. Liu, J. H. Huang, J. C. Sung and C. C. Lee, “Detection of cracks using
neural networks and computational mechanics”, Computer Methods in Applied
Mechanics Engineering, vol. 191, pp. 2831-2845, 2002.
182. Q. Chen, Y. W. Chan and K. Worden, “Structural fault diagnosis and isolation
using neural networks based on response-only data”, Computers and Structures,
vol. 81, pp. 2165-2172, 2003.
183. C. Y. Kao and S. L. Hung, “Detection of structural damage via free vibration
responses generated by approximating artificial neural networks”, Computers and
Structures, vol. 81, pp. 2631-2644, 2003.
184. M. Sahin and R. A. Shenoi, “Quantification and localisation of damage in beam-
like structures by using artificial neural networks with experimental validation”,
Engineering Structures, vol. 25, pp. 1785-1802, 2003.
185. S. Ataei, A. A. Aghakouchak, M. S. Marefat and S. Mohammadzadeh, “Sensor
fusion of a railway bridge load test using neural networks”, Expert Systems with
Applications, vol. 29, pp. 678-683, 2005.
186. J. Y. Kang, B. I. Choi, H. J. Lee, J. S. Kim and K. J. Kim, “Neural network
application in fatigue damage analysis under multiaxial random loadings”,
International Journal of Fatigue, vol. 28, pp. 132-140, 2006.
187. Y. J. Yan, L. Cheng, Z. Y. Wu and L. H. Yam, “Development in vibration-based
structural damage detection technique”, Mechanical Systems and Signal
Processing, vol. 21, pp. 2198-2211, 2007.
188. Z. X. Li and X. M. Yang, “Damage identification for beams using ANN based on
statistical property of structural responses”, Computers and Structures, vol. 86,
pp. 64-71, 2008.
189. M. Mehrjooa, N. Khaji, H. Moharrami and A. Bahreininejad, “Damage detection
of truss bridge joints using artificial neural networks”, Expert Systems with
Applications, vol. 35, pp. 1122-1131, 2008.
190. N. Bakhary, H. Hao and A. J. Deeks, “Structure damage detection using neural
network with multi-stage substructuring”, Advances in Structural Engineering,
vol. 13, no. 1, pp. 1-16, 2010.
191. A. H. Al-Rahmani, H. A. Rasheed and Y. Najjar, “A combined soft computing-
mechanics approach to inversely predict damage in bridges”, Procedia Computer
Science, vol. 8, pp. 461-466, 2012.
192. J. Shu, Z. Zhang, I. Gonzalez and R. Karoumi, “The application of a damage
detection method using Artificial Neural Network and train-induced vibrations on
a simplified railway bridge model”, Engineering Structures, vol. 52, pp. 408-421,
2013.
193. M. Yaghinin, M. M. Khoshraftar and M. Fallahi, “A hybrid algorithm for
artificial neural network training”, Engineering Applications of Artificial
Intelligence, vol. 26, pp. 293-301, 2013.
Reference
170
194. A. A. Elshafey, N. Dawood, H. Marzouk and M. Haddara, “Predicting of crack
spacing for concrete by using neural networks”, Engineering Failure Analysis,
vol. 31, pp. 344-359, 2013.
195. S. Erkaya, “Analysis of the vibration characteristics of an experimental
mechanical system using neural networks”, Journal of Vibration and Control,
vol. 18, no. 13, pp. 2059-2072, 2011.
196. P. C. Pandey and S. V. Barai, “Multilayer perceptron in damage detection of
bridge structures”, Computer & Structures, vol. 54, no. 4, pp. 597-608, 1995.
197. P. C. Karninsk, “The approximate location of damage through the analysis of
natural frequencies with artificial neural networks”, Journal of Process
Mechanical Engineering, vol. 209, pp. 117-223, 1995.
198. A. Seibi and S. M. Al-Alawi, “Prediction of fracture toughness using artificial
neural networks (ANNs)”, Engineering Fracture Mechanics, vol. 56, no. 3, pp.
311-319, 1997.
199. J. Zhao, J. N. Ivan and J. T. DeWolf, “Structural damage detection using artificial
neural networks”, Journal of Infra Structure Systems, vol. 4, pp. 93-101, 1998.
200. T. Marwala and H. E. M. Hunt, “Fault identification using finite element models
and neural networks”, Mechanical Systems and Signal Processing, vol. 13, no. 3,
pp. 475-490, 1999.
201. C. C. Chang, T. Y. P. Chang, Y. G. Xu and M. L. Wang, “Structural damage
detection using an iterative neural network”, Journal of intelligent material
systems and structures, vol. 11, pp. 31-42, 2000.
202. T. K. Lin, C. C. J. Lin and K. C. Chang, “A neural network based methodology
for estimating bridge damage after major earthquakes”, Journal of the Chinese
Institute of Engineers, vol. 25, no. 4, pp. 415-424, 2002.
203. C. Zang, M. I. Friswell and M. Imregun, “Structural damage detection using
Independent component analysis”, Structural Health Monitoring, vol. 3, no. 1,
pp. 69-83, 2004.
204. K. Worden and J. M. Dulieu-Barton, “Damage detection in bridges using neural
networks for pattern recognition of vibration signatures”, Engineering Structures,
vol. 27, pp. 685-698, 2005.
205. J. J. Lee, J. W. Lee, J. H. Yi, C. B. Yun and H. Y. Jung, “Neural networks-based
damage detection for bridges considering errors in baseline finite element
models”, Journal of Sound and Vibration, vol. 280, pp. 555-578, 2005.
206. N. Bakhary, H. Hao and A. J. Deeks, “Damage detection using artificial neural
network with consideration of uncertainties”, Engineering Structures, vol. 29, pp.
2806-2815, 2007.
207. F. S. Wong, M. P. Thint and A. T. Tung, “On-line detection of structural damage
using neural networks”, Civil Engineering Systems, vol. 14, pp. 167-197, 2007.
Reference
171
208. P. M. Pawar, K. V. Reddy and R. Ganguli, “Damage detection in beams using
spatial Fourier analysis and neural networks”, Journal of Intelligent Material
Systems and Structures, vol. 18, pp. 347-359, 2007.
209. C. Gonzalez-Perez and J. Valdes-Gonzalez, “Identification of structural damage
in a vehicular bridge using artificial neural networks”, Structural Health
Monitoring, vol. 10, no. 1, pp. 33-16, 2011.
210. P. Li, “Structural damage localization using probabilistic neural networks”,
Mathematical and Computer Modelling, vol. 54, pp. 965-969, 2011.
211. D. R. Parhi and A. K. Dash, “Application of neural networks and finite elements
for condition monitoring of structures”, Mechanical Engineering Science, vol.
225, pp. 1329-1339, 2011.
212. A. A. Elshafey, N. Dawood, H. Marzouk and M. Haddara, “Crack width in
concrete using artificial neural networks”, Engineering Structures, vol. 52, pp.
676-686, 2013.
213. O. Hasancebi and T. Dumlupınar, “Detailed load rating analyses of bridge
populations using nonlinear finite element models and artificial neural networks”,
Computers and Structures, vol. 128, pp. 48-63, 2013.
214. R. P. Bandara, H. T. T. Chan and P. D. Thambiratnam, “Structural damage
detection method using frequency response functions”, Structural Health
Monitoring, vol. 13, no. 4, pp. 418-429, 2014.
215. K. Aydin and O. Kisi, “Damage diagnosis in beam-like structures by artificial
neural networks”, Journal of Civil Engineering and Management, vol. 21, no. 5,
pp. 591-604, 2015.
216. S. S. Kourehli, “Damage assessment in structures using incomplete modal data
and artificial neural network”, International Journal of Structural Stability and
Dynamics, vol. 15, no. 6, p. 1450087, 2015.
217. H. A. Alavi, H. Hasni, N. Lajnef, K. Chatti and F. Faridazar, “An intelligent
structural damage detection approach based on self-powered wireless sensor
data”, Automation in Construction, vol. 62, pp. 24-44, 2016.
218. G. Yu, H. Qiu, D. Djurdjanovic and J. Lee, “Feature signature prediction of a
boring process using neural network modeling with confidence bounds”,
International Journal of Advance Manufacturing Technology, vol. 30, no. 7, pp.
614-621, 2006.
219. S. Y. Xiong and J. P. Withers, “An evaluation of recurrent neural network
modelling for the prediction of damage evolution during forming”, Journal of
Materials Processing Technology, vol. 170, pp. 551-562, 2005.
220. S. Ekici, S. Yildirim and M. Poyraz, “A transmission line fault locator based on
Elman recurrent networks”, Applied Soft Computing, vol. 9, pp. 341-347, 2009.
Reference
172
221. M. M. Abdelhameed and H. Darabi, “Neural network based design of fault-
tolerant controllers for automated sequential manufacturing systems”,
Mechatronics, vol. 19, pp. 705-714, 2009.
222. T. J. Connor, R. D. Martin and L. E. Atlas, “Recurrent neural networks and
robust time series prediction”, IEEE Transactions on Neural Networks, vol. 5, no.
2, pp. 240-254, 1994.
223. B. A. Pearlmutter, “Gradient calculations for dynamic recurrent neural networks:
A survey”, IEEE Transactions on Neural Networks, vol. 6, no. 5, pp. 1212-1228,
1995.
224. P. W. Tse and D. P. Atherton, “Prediction of machine deterioration using
vibration based fault trends and recurrent neural networks”, Journal of Vibration
and Acoustics, vol. 121, pp. 355-362, 1999.
225. C. G. Gan and D. Danai, “Model based recurrent neural network for modeling
nonlinear dynamic system”, IEEE Transactions of Systems, Man and
Cybernetics, vol. 30, no. 2, pp. 344-351, 2000.
226. Z. Waszczyszyn and L. Ziemianski, “Neural networks in mechanics of structures
and materials-new results and prospects applications, Computers and Structures,
vol. 79, pp. 2261-2276, 2001.
227. M. T. Valoor, K. Chandrasekhara and S. Agarwal, “Self-adaptive vibration
control of smart composite beams using recurrent neural network architectures”,
International Journal of Solids and Structures, vol. 38, pp. 7857-7874, 2001.
228. S. Seker, E. Ayaz and E. Turkcan, “Elman’s recurrent neural network
applications to condition monitoring in nuclear power plant and rotating
machinery”, Engineering Applications of Artificial Intelligence, vol. 16, pp. 647-
656, 2003.
229. A. M. Schafer and H. G. Zimmermann, “Recurrent neural networks are universal
approximators”, International Journal of Neural Systems, vol. 17, no. 4, pp. 253-
263, 2007.
230. A. Thammano and P. Ruxpakawong, “Nonlinear dynamic systems identification
using recurrent neural with multi-segment piecewise-linear connection weight”,
Memetic Computing, vol. 2, pp. 273-282, 2010.
231. R. Coban, “A context layered locally recurrent neural network for dynamic
system identification”, Engineering Applications of Artificial Intelligence, vol.
26, pp. 241-250, 2013.
232. H. Sohn, J. A. Czarnecki and C. R. Farrar, “ Structural health monitoring using
statistical process control”, Journal of Structural Engineering, vol. 126, no. 11,
pp. 1356-1363, 2000.
233. M. L. Fugate, H. Sohn and C. Farrar, “Vibration-based damage detection using
statistical process control”, Mechanical Systems and Signal Processing, vol. 15,
no. 4, pp. 707-721, 2001.
Reference
173
234. Y. Lei, A. S. Kiremidjian, K. K. Nair, J. P. Lynch, K. H. Law, T. W. Kenny, E.
Carryer and V. Kottapalli, “Statistical damage detection using time series analysis
on a structural health monitoring benchmark problem”, Proceedings of the 9th
International Conference on Applications of Statistics and Probability in Civil
Engineering, San Francisco, pp. 6-9, 2003.
235. Y. Lu and F. Gao, “A novel time-domain auto-regressive model for structural
damage diagnosis”, Journal of Sound and Vibration, vol. 283, pp. 1031-1049,
2005.
236. S. G. Mattson and S. M. Pandit, “Statistical moments of autoregressive model
residuals for damage localisation”, Mechanical Systems and Signal Processing,
vol. 20, pp. 627-645, 2006.
237. J. Zhang, Y. L. Xu, Y. Xia and J. Li, “A new statistical moment-based structural
damage detection method”, Structural Engineering and Mechanics, vol. 30, no. 4,
pp. 445-466, 2008.
238. M. Gul and F. N. Catbas, “Statistical pattern recognition for Structural Health
Monitoring using time series modeling: Theory and experimental verifications”,
Mechanical Systems and Signal Processing, vol. 23, pp. 2192-2204, 2009.
239. S. Law and J. Li, “Updating the reliability of a concrete bridge structure based on
condition assessment with uncertainties”, Engineering Structures, vol. 32, pp.
286-296, 2010.
240. J. L. Zapico-Valle, M. Garcia-Dieguez, M. P. Gonzalez-Martınez and K. Worden,
“Experimental validation of a new statistical process control feature for damage
detection”, Mechanical Systems and Signal Processing, vol. 25, pp. 2513-2525,
2011.
241. K. Kwon and D. M. Frangopol, “Bridge fatigue assessment and management
using reliability-based crack growth and probability of detection models”,
Probabilistic Engineering Mechanics, vol. 26, pp. 471-480, 2011.
242. A. A. Mosavi, D. S. Dickey and S. Rizkalla, “ Identifying damage locations under
ambient vibrations utilizing vector autoregressive models and Mahalanobis
distances”, Mechanical System and Signal Processing, vol. 26, pp. 254-267,
2012.
243. F. Cavadas, I. F. C. Smith and J. Figueiras, “Damage detection using data-driven
methods applied to moving-load responses”, Mechanical Systems and Signal
Processing, vol. 39, pp. 409-425, 2013.
244. F. P. Kopsaftopoulos and S. D. Fassois, “A functional model based statistical
time series method for vibration based damage detection, localization, and
magnitude estimation”, Mechanical Systems and Signal Processing, vol. 39, pp.
143-161, 2013.
Reference
174
245. B. Phares, P. Lu, T. Wipf, L. Greimann and J. Seo, “Field validation of a
statistical-based bridge damage-detection algorithm”, Journal of Bridge
Engineering, vol. 18, no. 11, pp. 1227-1238, 2013.
246. D. Wang, W. Xiang and H. Zhu, “ Damage identification in beam type structures
based on statistical moment using two step method”, Journal of Sound and
Vibration, vol. 333, pp. 745-760, 2014.
247. L. Yu and J. Zhu, “Nonlinear damage detection using higher statistical moments
of structural responses”, Structural Engineering and Mechanics, vol. 54, no. 2,
pp. 221-237, 2015.
248. A. J. Reiff, M. Sanayei and R. M. Vogel, “Statistical bridge damage detection
using girder distribution factors”, Engineering Structures, vol. 109, pp. 139-151,
2016.
249. M. G. Masciottaa, L. F. Ramos, P. B. Lourenco, M. Vastab and G. D. Roeck, “A
spectrum-driven damage identification technique: Application and validation
through the numerical simulation of the Z24 Bridge”, Mechanical Systems and
Signal Processing, vol. 70, pp. 578-600, 2016.
250. M. Dilenaa, M. P. Limongelli and A. Morassi, “Damage localization in bridges
via the FRF interpolation method”, Mechanical Systems and Signal Processing,
vol. 52, pp. 162-180, 2015.
251. H. Gokdag, “A crack identification method for beam type structures subject to
moving vehicle using particle swarm optimization”, Gazi University Journal of
Science, vol. 26, no. 3, pp. 439-448, 2013.
252. F. Kang, J. Li and Q. Xu, “Damage detection based on improved particle swarm
optimization using vibration data”, Applied Soft Computing, vol. 12, pp. 2329-
2335, 2012.
253. D. Hester and A. Gonzalez, “A wavelet-based damage detection algorithm based
on bridge acceleration response to a vehicle”, Mechanical Systems and Signal
Processing, vol. 28, pp. 145-166, 2012.
254. B. Sahoo and D. Maity, “Damage assessment of structures using hybrid neuro-
genetic algorithm”, Applied Soft Computing, vol. 7, pp. 89-104, 2007.
255. G. Li, S. He, Y. Ju and K. Du, “Long distance precision inspection method for
bridge cracks with image processing”, Automation in Construction, vol. 41, pp.
83-95, 2014.
256. R. S. Adhikari, O. Moselhi and A. Bagchi, “Image-based retrieval of concrete
crack properties for bridge inspection”, Automation in Construction, vol. 39, pp.
180-194, 2014.
257. J. Li and H. Hao, “Health monitoring of joint in steel truss bridges with relative
displacement sensors”, Measurement, vol. 88, pp. 360-371, 2016.
258. W. T. Thomson, “Theory of vibration with application”, Third Edition, CBS
Publishers, New Delhi, India.
Reference
175
259. L. Fryba, “Vibration of solids and structures”, Third Edition, Thomas Telford
Ltd, Prague, 2002.
260. M. Paz, “Structural Dynamics: Theory and Computation”, Third Edition, An
International Thomson Publishing Company, New York.
261. A. K. Chopra, “Dynamics of structures: Theory and application to earthquake
engineering”, Fourth Edition, Prentice-Hall International Series, New York.
262. D. C. Montgomery, “Introduction to statistical control”, Fifth edition, Wiley
Publications.
263. P. J. Brockwell and R. A. Davis, “Time series: Theory and methods”, Second
edition, Springer, 1999.
264. G. E. P. Box, G. M. Jenkins and G. C. Reinsel, “Time series analysis: Forecasting
and control”, Third edition, Prentice Hall International, 1994.
265. H. Demuth and M. Beale, “Neural Networks Toolbox”, The Math Works, Inc
Natick, 2002.
266. H. Yu and B. M. Wilamski, “Levenberg-Marquardt Training, Lecture notes on
intelligent systems, Auburn University.
176
Dissemination
Publications:
International Journals
1) Vibrational Analysis of Structures: A Review, Dayal. R Parhi and Shakti. P Jena,
International Journal of Applied Artificial Intelligence Engineering Systems, vol. 4, no. 2,
pp.83-94, 2012.
2) Dynamic Response of Cracked Simple Beam Carrying Moving Mass, Shakti. P Jena and
Dayal. R. Parhi, International Journal of Applied Engineering Research, vol. 9, no. 26,
pp.8794-8797, 2014. (Scopus)
3) Dynamic Deflection of a Cantilever Beam Carrying Moving Mass, Shakti. P Jena and
Dayal. R. Parhi, Applied Mechanics and Materials, vols. 592-594, pp. 1040-1044, 2014.
(Scopus)
4) Dynamic Response of Damaged Cantilever Beam Subjected to Traversing Mass,
International Journal for Technological Research in Engineering, vol. 2, no. 7, pp. 860-86,
2015.
5) Parametric Study on the Response of Cracked Structure Subjected to Moving Mass,
Shakti. P Jena and Dayal. R. Parhi, Journal of Vibration Engineering and Technology
(Accepted to be published in vol. 5, no. 1, 2017. (SCI)
6) Comparative Study on Cracked Beam with Different types of cracks Carrying Moving
Mass, Shakti. P Jena, Dayal R. Parhi and Devasis Mishra, Structural Engineering and
Mechanics, International Journal, vol. 56, no. 5, pp.797-811, 2015. (SCI)
7) Dynamic and experimental analysis on response of multiple cracked structures carrying
transit mass, Dayal. R Parhi and Shakti. P Jena, Journal of Risk and Reliability, SCI
(Accepted)
8) Dynamic Response and Analysis of Damaged Beam Structure Subjected to Traversing
Mass, Shakti. P Jena and Dayal R. Parhi, Journal of Steel and Composite Structures.
(Under review since 8th April 2016, SCI)
9) Response Analysis of Cracked Structure Subjected to Transit Mass- A Parametric
Study, Shakti. P Jena and Dayal R. Parhi, Journal of vibroengineering. (Under review
since 18th April 2016, SCI)
10) Response of structure to high speed mass, Shakti. P Jena and Dayal R. Parhi, Procedia
engineering, vol. 144, pp.1435-1442, 2016. (Scopus)
Dissemination
177
International Conferences
1) Analytical and Computational Study for the Dynamic Response of a Cantilever Beam
Carrying Moving Mass, Shakti. P Jena and Dayal R. Parhi, International Conferences in
Smart Technologies for Mechanical Engineering, 25-26 October, 2013, Delhi
Technological University, Delhi.
2) Dynamic Analysis of Cantilever beam with Moving Mass, Shakti. P Jena and Dayal R.
Parhi, International Conference on Structural Engineering and Mechanics, 20-22
December, 2013, NIT, Rourkela, Odisha.
3) Dynamic response of a simply supported beam with traversing mass, Shakti. P Jena and
Dayal R. Parhi, International Conference on Industrial Engineering Science and
Applications, 2-4 April, 2014, NIT, Durgapur, Westbengal.
4) Numerical Analysis for the Dynamic Analysis of Cantilever Beam Carrying Moving Load,
Shakti. P Jena and Dayal R. Parhi, Sixth International Conference on Theoretical,
Applied, Computational and Experimental Mechanics, 29-31 December, 2014, IIT,
Kharagpur, Westbengal.
5) Response of Cracked Cantilever Beam Subjected to Traversing Mass, Shakti. P Jena,
Dayal R. Parhi and Devasis Mishra, 2-3 December, 2015, ASME India Gas Turbine
Conference, HICC, Hyderabad, Andhrapradesh.
178
Vitae
Mr Shakti Prasanna Jena was born in 1st
July, 1984 in Jajpur, Odisha. He has completed
his +2 Science education from Kendrapara College (Kendrapara), in 2001 and Bachelor of
Engineering in Mechanical Engineering from Orissa Engineering College, (B.P.U.T.
University) in 2007. He got his Master of Technology in Mechanical Engineering
(Mechanical System Design) from B.P.U.T (Rourkela) in 2011. After completion of his
Post-Graduation, he joined the Ph.D. program at Department of Mechanical Engineering,
National Institute of Technology, Rourkela, in July 2012 and submitted her Ph.D. thesis in
July 2016. His research interest includes mechanical vibration, structural dynamics,
damage analysis in cracked structure and various soft computing methods etc.